the personal earnings distribution: individual and institutional determinants. final report for the
TRANSCRIPT
ED 105 099
AUTHORTITLE
INSTITUTION
SiJNS AGENCY
REPORT NOPUB DATENOTEAVAILABLE FROM
EDRS PRICEDESCRIPTORS
DOCUMENT RESUME
CE 003 348
Bluestone, Barry A.The Personal Earnings Distribution: Individual andInstitutional Determinants. Final Report for thePeriod July 1, 1970--October 31, 1973.Michigan Univ., Ann Arbor. Inst. of Labor andIndustrial Relations.Manpower Administration (DOL), Washington, D.C.Office of Research and Development.DLMA 91-24-70-51-1Sep 74432p.; Ph.D. Dissertation, University of MichiganNational Technical Information Service, Springfield,Virginia 22151 ($6.00)
HF-$0.76 HC-$22.21 PLUS POSTAGEDoctoral Theses; *Economic Research; Employees;Employment Problems; Human Resources; IndustrialPersonnel; Industry; Labor Force; Labor Market; LaborUnions; *Low Income; Manpower Utilization; MultipleRegression Analysis; Occupational Mobility; *ResearchMethodology; *Salary Differentials; *Theories;Unskilled Workers; Wages
ABSTRACTThe study investigates the determinants of the
earnings distribution in the U.S. paying particular attention to theless-skilled segment of the workforce. A general earnings theory isproposed which has elements of human capital theory, institutionalhypotheses, and radical stratification analysis. Much attention ispaid to testing the "crowding" hypothesis that workers restricted toemployment in A limited number of industries or occupations will bepaid substantially less than workers who are not so restricted. Theregression results, based on a large integrated micro-macro data set,yield extensive evidence of stratification and industry variablesaffecting earnings after controlling for differences in humancapital. This is especially true among less-skilled workers. Theresearch includes both a literature review and extensivebibliography. (Author)
.f.
Report DLMA 91-24-70-51-1
THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL
AND INSTITUTIONAL DETERMINANTS
Barry A. Bluestone
Institute of Labor and Industrial RelationsUniversity of Michigan-Wayne State UniversityAnn Arbor, Michigan 48104
Septe.per 1974
U S DEPARTMENT OF SEALS'SEDUCATION & WELFARENA TIC- 'I- INSTITUTE OF
EDUCATIONDOCUMENT r+AS BEEN REPRO
OuCE 0 EX:.; I, Y CI' r.I O I ROM..E Pf RSON ON ORIGIN
,NG IT POINT :JI 4rt 4..,OR OPINIONSSi .5110 )O NOT MCI y RE PRE.
st-ht or FICos, NAT.,)NA. rNs*Iitif I 0(C)0C0I.ON Ps, not. OR POI +EY
Final Report for Period July 1, 1970 - October 31, 1973
Prepared for
U. S. Department of LaborManpower AdministrationOffice of Research and Development
Washington, D.C. 20210
LI) 2CJ
Report DLMA 91-24-70-51-1
THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL
AND INSTITUTIONAL DETERMINANTS
Barry A. Bluestone
Institute of Labor and Industrial Relations
University of Michigan-Wayne State University
Ann Arbor, Michigan 48104
September 1974
Final Report for Period July 1, 1970 - October 31, 1973
Prepared for
U. S. Department of Labor
Manpower AdministrationOffice of Research and Development
Washington, D.C. 20210
This report was prepared for the Manpower Administration,
U.S. Department of Labor, under research and development
grant no. DL 91-24-70-51. Since grantees conducting
research and development projects under Government sponsor-
ship are encouraged to express their own judgement freely,
this report does not necessarily represent the official
policy of the Department of Labor. The grantee is solely
responsible for the contents of this report.
t -! a
STANDARD TITLE PACE 1.
FOR TCC11N:CAL PrPORTS 9 I -?11-70-51- 1
3. !(....yo_ ( at slx No.
4. fide 3 Hcpo:: Dat..The Personal Earnings Distributioo: Individual Sep, 74.
and Institutional Detenni,lants 6. l'ericoseteg Organ...mh, cosh.
7. Aul.kot(.)
Rarry_Blo-stonc9. Petto:gaint, thtz.>niZat:ti Ad.!:
Jnstitute of Labor and 'Industrial RelationsUniversity of Michigan- Wayne State UniversityAnn Arbor, Michigan 48104
8. P:No:nu:à; 0:ga:4/atom Kept.No.
i0. Plow 1 o:k (hilt No.
11. Co:lac: ( :art No.
DL 9l -24 -70 -51
12. SpOrl.;A: :re 4.era, s N.U.S. DerarM,:ent of LnborManpower Ad:Anistra.,:ion
Office of'c'search and Dcvelolcment
1111 20th St., N.W. Washington, D.C. 20210
15. Supple me mar y
13. l)peofRepo:t&I-iod
Final
14. Spowsoling Agency Code
16. /.1.1:act,This study invcstivites the determinants of the earnings distribution
in the U,S, paying particular attention to the less-skilled segment ofthe workforce. A general earnincs theory is proposed which has elementshumaa capital theory, institutional hypothescs, cadical stratification
analysis. Much attention is paid to testing the "crowding" hypothesisthat workers restricted to employment in a limited number of industriesor occupations will be paid substantially less than workers who are not sorestricted, The regression results, based on a large integrated micro-macro data sct, yield extensive evidence of stratification and industryvariables affecting earnings i.fter controlling for differences in human
capital. This is especially true among less-skilled workers, The research
includes both a literature review and extensive bibliography.
17. Key lord`; Lio....raen: 1/u. I/est :sptor..c
Earnings; Employment; Labor; Nanpower; Unionization
17b. kentificfs/Open-Ende'd 'terms The Labor MarketLabor Force, Labor Market, and Labor RequirementsLabor Market Processes
17c- COCA Ti hiK:our 5C
18. Da:At:m:01 `tate:,ntDistribution is unlimited. Available from
Nation,.1 Teehrical Information Service, SpringfieldVa., 22151,
Faroe CFSTI-35 .47e>
19. :,ecuricy Class (Thl.Report)
17NCI-A"`ii'iLB26: :Net Of It (I hi
1).11).ITN( I Aqs11:11'1)
21. No. oi429pp.
22.
$6.001...:(.01mq DC ,50051 I
THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL
AND INSTITUTIONAL DETERMINANTS
by
Barry Alan Bluestone
A dissertation submitted in partial fulfillmentof the requirements for the degree of
Doctor of Philosophy inThe University of Michigan
1974
Doctoral Committee:
Professor Harold Levinson, ChairmanProfessor Daniel FusfeldAssistant Professor Malcolm CohenProfessor Gerald Gurin
ABSTRACT
THE PERSONAL EARNINGS DISTRIBUTION: INDIVIDUAL
AND INSTITUTIONAL DETERMINANTS
by
Barry Alan Bluestone
Chairman: Professor Harold Levinson
This study investigates the determinants of the earnings
distribution in the United States paying particular attention to the
less-skilled segment of the workforce.
A general earnings theory is proposed which has elements of
human capital theory, institutional hypotheses, and radical strati-
fication analysis. Much attention is paid t, testing the "crowding"
hypothesis that workers restricted to employment in a limited number
of industries or occupations will be paid substantially less than
workers who are not so restricted. It was hypothesized that after
controlling for differences in human capital, large wage differentials
would continue to exist for similarly qualified workers. These
differences could be attributed to the stratification of the labor
force, particularly by race and sex. Once stratified, differences in
industry characteristics would have an effect on the personal earnings
2
distribution as well. Those workers who gain employment in the
more concentrated, profitable, and unionized industries will earn
more than others who have similar work characteristics.
The regression results, based on a large integrated miclo-
macro data set, yield extensive evidence of stratification and industry
variables affecting earnings after controlling for differences in
human capital. This is especially true among the least skilled
workers in the labor force although there'is a substantial earnings
effect throughout most of the occupational hierarchy. While it was
impossible to obtain incontrovertible evidence that "crowding" was the
culprit in producing "human capital constant" wage differentials,
the evidence seems to point overwhelmingly in this direction.
Concentration and unionization also have a significant impact on
wages as well as a number of other industry factors.
The overriding policy implication following from this analysis
is that large scale government intervention is required in order to
correct the apparently massive "inefficiencies" that currently exist
in American labor markets. Intervention is required to equalize
human capital investment opportunities but equally important to break
down the barriers to inter-occupational and inter-industry mobility
that apparently still exist.
7
C) Barry Alan Bluestone 1974
All Rights Reserved
Reproduction by the United StatesGovernment in whole or in part ispermitted for any purpose.
The material in this project was prepared under
Grant No. 91-24-70-51 from the Manpower Administration,
U.S. Department of Labor; under the authority of Title I
of the Manpower Development and Training Act of 1962.
Researchers undertaking such projects under Government
sponsorship are encouraged to express freely their
professional judgment. Therefore, points of view or
opinions stated in this document do not necessarily
represent the official position or policy of the
Department of Labor.
iii
9
PREFACE
The present study began a number of years ago, when in the
course of poverty research, the common stereotype of the poor was
shattered by the revelation that the majority of the poor work
and in nearly a third of all poor households the head works full-time
all year round.1
'
2 The AFDC mother, the aged, the infirm, the
industrially "undisciplined," in short those out of the labor force
or unemployed were found to be only a portion of the poor. For many
others poverty was discovered to be the result of low-wage employment
rather than no employment at all.
Many of the particular causes which explain the poverty of the
nonworking poor--sickness, old age, illiteracy, and "bad luck"--fail
to adequately explain the poverty incomes of those who work. For
them poverty is a much more complex phenomenon going beyond individual
1Some of the earliest research on the working poor include:
George Delehanty and Robert Evdns, Jr., "Low-Wage Employment: An
Inventory and an Assessment" (Northwestern University, n.d.)
unpublished manuscript; Laurie D. Cummings, "The Employed Poor:
Their Characteristics and Occupations," Monthly Labor Review, July
1965; Dawn Wachtel, The Working Poor (Ann Arbor: Institute of Labor
and Industrial Relations, University of Michigan-Wayne State
University, 1967) mimeo; Barry Bluestone, "Low-Wage Industries
and the Working Poor," Poverty and Human Resources Abstracts, March
1968.
2Computed from "Work Experience of Family Heads, by Poverty
Status of Family, 1968," U.S. Department of Labor, Manpower Report
of the President (Washington: U.S. Government Printing Office),
March 1970, Table 1, p. 121.
iv
10
inadequacies. The confluence of market forces and personal
attributes forms a complex web from which the individual factors
contributing to low earnings are difficult to unravel. Wage theory
should help us understand the problem, but so far it has generally
failed.
The reason for this is.that the simplifying assumptions in
traditional wage theory tend to confuse the issue. The assumption of
a homogeneous labor force, found in the institutionalist framework,
tends to obscure the earnings effect of differences in skills and
competencies among workers. Alternatively assuming perfect competi
tion and labor mobility as in the pure human capital theory obscures
many other factors which impose their own order on the distribution
of income and earnings. An understanding of the working poor
requires a general wage theory that focuses on both the characteristics
of workers themselves, and on the labor markets in which they work,
while dropping the restrictive assumptions normally found in
traditional wage theory. To understand the determinants of low
wages requires an understanding of the whole distribution of earnings.
What began as a narrow study of poverty employment thus blossomed
into a more general investigation of the determinants of personal
earnings in the United States.
My original interest in the problem was spurred by Louis Ferman,
Director of the Research Division of the Institute of Labor and
Industrial Relations, University of Michigan-Wayne State University.
My colleagues Mary Stevenson and Charles Betsey helped prepare the
v
11
data set and contributed to some of the analysis. William Murphy
was indispensible in writing computer programs well beyond my
capability while Lynn Ware, James Sumrall, Jr., and Martha
MacDonald troubled over some of the theory and mathematical presenta-
tion with me. Countless friends associated with the Union for
Radical Political Economics were helpful at various times in
suggesting hypotheses to test and always kept steering my research
in relevant directions. Mrs. Kathleen Schwartz was responsible for
diligently typing the final draft. Finally a special note of
appreciation goes to my dissertation committee, Professors Harold
Levinson, Malcolm Cohen, Daniel Fusfeld, and Gerald Gurin. The
committee aided me immeasurably in the development of the project and
always did their best to force me to consider all sides of the issues
involved. To all of these people I extend my warmest appreciation for
their help and their friendship.
vi
12
CONTENTS
LIST OF TABLES
Page
.. ix
LIST OF ILLUSTRATIONS xii
Chapter
I. INTRODUCTION 1
Wage Theory and the Study of PovertyThe Design of the Study
II. EXISTING THEORIES OF WAGE DETERMINATION 11
Marginal Productivity TheoryHuman Capital TheoryThe Basic Human Capital ModelEmpirical Studies of the Human Capital
Earnings FunctionThe Institutional ApproachBalkanized Labor MarketsAn Institutional Model of Wage DeterminationEmpirical Verification of the Institutional
ApproachProblems with the Institutional ApproachSocial Stratification AnalysisSocial Stratificatioa and Wage Theory
Toward a Complete Theory of Wage Determination
III. A SOCIAL STRATIFICATION Tr*',,,RY OF PERSONAL
EARNINGS 64
The General Stratification ModelThe Specific ModelA Mathematical TreatmentThe Reduced FormThe Data Source and the Set of Regression Variables
IV. THE ESTIMATION PROCEDURE 111
Potential MulticollinearityThe Estimation Procedure MechanicsProblems in Parameter Specification
vii
13
CONTENTS (Continued)
Chapter Page
V. THE REGRESSION RESULTS 128
The Regression ResultsOccupation Stratum 1-3Occupation Stratum 5Occupation Stratum 6-9Occupation Stratum 12-14Occupation Stratum 15-17Cross Occupation StrataThe "Grand Pooled" Regressions
VI. AN EVALUATION OF THE REGRESSION RESULTS 223
The MethodologyThe Z* MeasureThe ResultsThe Relative Impact of Human Capital and
Non-human Capital Factors
VII. CONCLUSIONS AND IMPLICATIONS 274
Theories to Explain these ResultsWhat's To Be Done?
APPENDICES
A. THE DATA SOURCE 295
The SEO Data BaseConstruction of Occupation Scores
Occupation DataConstruction of Industry Data Base
Industry Data
B. MEANS, STANDARD DEVIATIONS, AITCORRELATION MATRICES 334
BIBLIOGRAPHY 405
viii
14
LIST OF TABLES
TablePage
4.1 Zero-Order Correlation Matrix- -White Males/
All Occupation Strata 114
5.1 Regression Equations: Occupation Stratum 1-3
by Race and Sex 135
5.2 Regression Coefficients on the Stratification
Variables in the Pooled OCC STRATUM 1-3 Regressions 146
5.3 Regression Equations: Occupation Stratum 5 by
Race and Sex 150
5.4 Partial (XtX) Matrix for Occupation Stratum 5 160
5.5 Regression Equations: Occupation Stratum 6-9 by
Race and Sex 164
5.6 Regression Equations: Occupation Stratum 12-14
by Race and Sex
5.7 Regression Equations: Occupation Stratum 15-17
by Race and Sex
174
186
5.8 Regression Equations: All Occupation Strata
by Race and Sex 191
5.9 Potential Wage Rates for Black Male Workers
Under Varying Assumptions about their
Characteristics 201
5.10 Potential Wage Rates for White Female Workers
Under Varying Assumptions about their
Characteristics 207
5.11 Potential Wage Rates for Black Female Workers
Under Varying Assumptions about their
Characteristics 210
!
ix
is
LIST OF TABLES (Continued)
Table Page
5.12a Estimated Hourly Earnings under VariousAssumptions Concerning the Human Capital
Module in the Occupation-Pooled Regressions:School Completed = 8 Years 213
5.12b Estimated Hourly Earnings under VariousAssumptions Concerning the Human CapitalModule in the Occupation-Pooled Regressions:School Completed = 12 Years 214
5.12c Estimated Hourly Earnings under VariousAssumptions Concerning the Human Capital
Module in the Occupation-Pooled Regressions:School Completed = 16 Years 215
5.13 Partial (XtX) Matrix for "Grand" PooledRegression 220
5.14 Summary of Significant %MININD Factors 221
6.1 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum- -
White Males 238
6.2 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum- -
Black Males 248
6.3 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum- -
White Females 253
6.4 Wage Intervals due to Stratification andIndustry Factors, By Occupation Stratum-Black Females 257
6.5 Pooled Occupation Regression WageInterval Estimates 259
6.6 Wage Intervals due to Stratification endIndustry Factors, By Occupation Stratum- -
All Race-Sex Groups 260
6.7 Z/HC Ratios By Occupation Strata 267
x
16
LIST OF TABLES (Continued)
Table Page
A.1 The SEO Sample Population of FullTimeFull Year Workers 300
A.2 General Education Development 306
A.3 Specific Vocational Preparation Scale... 307
A.4 The Physical Demands Scale for Strength 307
AS "Occupation Levels" based on GED and
SVP Scores 310
A.6 Census Occupations by "Occupation Level" 312
A.7 Occupation Data by SEO Occupation Code 320
A.8 The Industry Data 331
I
xi
17
LIST OF ILLUSTRATIONS
Figure
2.1 Marginal Revenue Products with Labor Supply
Page
Restrictions 20
2.2 The Institutionalized Wage Bargain 41
3.1 A Social Stratification Model of the Personal
Earnings Distribution 71
3.2 No "Crowding" 74
3.3 Simple "Crowding" 75
3.4 Complex "Crowding" 76
4.1 The Theoretical Relationship betweenEarnings and Experience 123
xii
18
CHAPTER I
INTRODUCTION
What factors determine an individual's wage? How are earnings
related to "skills" and to "productivity." To what extent does the
distribution of personal earnings reflect the distribution of skills
and to what extent institutional factors? These are the fundamental
questions which are the concern of this dissertation.
There exists today no generally agreed upon wage theory.
Rather there exists a set of hypotheses, each constrained by its own
set of assumptions, each with its own distinct set of "exogeneous
variables," and each in competition with the other. Consequently,
there is general confusion over the relationship between "wage,"
"skill," and "productivity." Adam Smith's theory of "compensatory"
wage differentials, J.B. Clark's marginal productivity theory, and
the investment theory of earnings stemming from the work of Denison,
Schultz, and Becker all posit a tight relationship between an
individual's own skills, productivity, and earnings. In opposition,
institutionalist wage theory discards the neoclassical assumptions
of perfect product and labor markets thereby disrupting a more
perfect mapping of human capital into the distribution of earnings.
Industrial characteristics replace human capital attributes as the
primary variables in institutional wage theory. "Radical" wage
1
19
2
theory, based on social stratificationanalysis, goes beyond the
institutionalist critique, severing any link between personal human
capital investment decisions and individual earnings. An individual's
stock of human capital, according to radical theory, is a function
of class, race, and sex. Relative earnings are determined by social
status rather than individual choice in the acquisition of human
capital.
The competition between neoclassical, institutionalist, and
radical theories remains largely unresolved.1 Each theory has a
distinct wage generating function which explains only a portion of
the existing variance in earnings. Generally the theories have not
been tested against each other. Consequently, a wage theory synthesis
has not evolved, much less a new scientific "paradigm," to use the
terminology of Thomas Kuhn.2
Yet the framework for a synthesis can be constructed. By
substituting the assumptions of institutionalist and radical theory
into the overly restrictive model of the neoclassical paradigm, a
"flexible" general wage theory can be developed. Specifically
accounting for imperfect product and labor markets produces a wage
theory capable of defining the complex links between human capital,
industry and occupational attachment, and the distribution of earnings.
1For the best discussion of the competition, see David M.
Gordon, Theories of Poverty and Underemployment (Lexington: D.C. Heath,
1972).
2Thomas Kuhn, The Structure of Scientific Revolutions (Chicago:
University of Chicago Press, 1962).
20
3
In the general theory developed here a human capital earnings
function is modified to allow for institutional barriers to
industrial and occupational access along racial and sexual lines.
The independent effects of industry characteristics such as unioni-
zation, concentration, profits, and capital-intensity also enter into
the wage model.
Wage Theory and the Study of Poverty
A correct specification of the wage determination process is of
more than academic interest. A good ?art of the government's
antipoverty strategy of the 1960s was based on labor market studies.
Translated into public policy, human capital research contributed to
the emphasis on manpower and human resource development programs.
Along this line, the so-called war on poverty was designed to "find
new means for offering disadvantaged groups in urban and rural
America a chance to develop their own capacities and become productive
members of our society.0 Federal outlays for all manpower
activities rose steadily during the latter part of the decade in
response to the presumption "that education and training are
specially effective ways to bring people out of poverty."4 Programs
totalling only $184 million dollars in 1964 grew to nearly $2.4 billion
3,'The Budget Message of t
Printing Office, 1969), p. 47.
he President," The Budget of
States Government, Fiscal Year
4Thomas I. Ribich,
Institution, 1968), p. 1.
1970 (Washington, D.C.: U.S.
Educationion and Poverty (Washington:
21
the United
Government
Brookings
4
by 1970.5
The results of this effort were mixed. The government's
attempt to "upgrade," "rehabilitate," "train," "retrain," "integrate,"
"reintegrate," and "prepare" the poor for the job market was often in
vain. The payoff in terms of employment gains and increasing earnings
frequently failed to live up to expectations. No matter how measured,
the cost of a particular prpgram often exceeded the benefits. Many
programs had low job retention rates and often entrants did not
complete their training cycle. In other cases trainees completed a
manpower program only to find it impossible to gain adequate employ-
6ment.
Particular manpower programs failed because of insufficient
funding, lack of coordination, inadequate training, and poor
forecasting of job opportunities. But even the successes brought
little reason for enthusiasm. For those who completed MDTA training
in the middle of the 1960s, only three out of five advanced in pay,
and the increased earnings were quite small. According to the largest
study of MDTA, involving over 100,000 institutional training graduates,
5Sar A. Levitan, "Manpower Programs Under Republican
Management," Poverty and Human Resources, March-April, 1970, p. 12.
6In a comprehensive analysis of the institutional portion of
the MDTA program it was found that over 30 percent of the trainees
dropped out before completing vocational training and only 58 percent
found jobs related to their training curriculum. For a comprehensive
overview of the manpower programs in the 1960s, see Sar A. Levitan
and Garth L. Mangum, Federal Training and Work Programs in the Sixties
(Ann Arbor: Institute of Labor and Industrial Relations, 1969).
22
5
the average wage for males after training was $2.06 an hour,
27 percent higher than the average pretraining wage. For females,
the post-training wage was raised to $1.53, less than 20 percent
above pretraining levels and below the then prevailing federal minimum
wage.7 What is worse, these statistics apply only to those who
actually completed a manpower program and found jobs. Thousands of
other failed to complete programs and others finished and could find
no suitable employment.8
What explains the apparently low returns on investment in
manpower programs? One explanation, of course, is that existing
programs actually add little to the "human capital" of the disadvantaged.
Much more extensive human resource programs are necessary before
satisfactory returns can be anticipated. The other explanation rests
on the hypothesis that a lack of human capital is not the major barrier
to economic success for the poor. Augmenting an individual's stock
7U.S. Department of Labor, Manpower Administration, "The
Influence of MDTA on Earnings," Manpower Evaluation Report, No. 8
(Washington: U.S. Government Printing Office, December 1968), p. 18.
8In place of institutional manpower programs, on-the-job training
funded by the federal government has provided more people directly
with jobs. But according to a GAO report, the federally-funded on-the-
job training program is no more than a transfer system whereby the
government pays for specific job training which would normally be pro-
vided by the cooperating firm in spite of the program. The General
Accounting Office uncovered the fact that: "OJT contracts had served
primarily to reimburse employers for OJT which they would have con-
ducted even without the government's financial assistance. These
contracts were awarded even though the intent of the program was to
induce new or additional training efforts beyond those usually carried
out." See U.S. General Accounting Office, "Improvements Needed in
Contracted for On-the-Job Training under the Manpower Development and
Training Act" (Washington: U.S. Government Printing Office, 1968).
23
6
of human capital, it may be argued, yields an insignificant marginal
return because employment opportunities are nonexistent or highly
restricted. The "low-wage" workforce may possess the human capital
characteristics of higher paid members of the labor force, but fail
to earn larger incomes because of geographical immobility, the high
cost of job information, or racial and sexual discrimination. Low
relative earnings can result from the "crowding" of a subset of the
workforce into a limited number of industries and occupations.9
Denied access to other economic sectors for which they are qualified
on the basis of human capital, members of the "low-wage" workforce may
be competing with each other for the limited supply of jobs in the
sectors open to them. In this case, the maintenance of an "oversupply"
of workers in the "low-wage" sector may be the primary reason for low
wages, not a lack of human capital. In addition, the industries to
which economic minorities are restricted may consist of marginal firms.
operating within an economic environment characterized by low capital-
labor ratios, strong product market competition, and weak unions.
For any given degree of "crowding," firms in less "permissive"
economic environments may offer lower wages. Poverty earnings will
then be a function of social "underemployment" rather than personal
"underinvestment."
"Relative Underemployment" can be said to exist when an individual
qualifies for higher wage employment on the basis of human capital but
9The "crowding" hypothesis can be traced to F.Y. Edgeworth,"Equal Pay to Men and Women," Economic Journal, December 1922.
L 24
7
is denied access to it on other grounds. If underemployment is
widespread among low-wage workers, the problem of low-wage employment
is then only partly the effect of inadequate human capital. In this
case manpower programs will have a limited ability to make improvements
in the earnings of the low-paid.
To what extent low wages reflect inadequate amounts of human
capital versus restricted access to employment opportunity can only
be ascertained through an empirical investigation which permits both
factors to simultaneously enter the analysis. This is the reason
for developing a testable "general" wage theory. Measuring the effect
of human capital on the wage rate relative to the effects of industry
variables and restricted employment opportunity is the necessary
prerequisite for understanding both the promise and the limitations
of manpower policy. Beyond this, the testing of a general wage theory
focuses attention on the factors which are most important in the wage
determination process for all members of the workforce. Such research
can empirically account for the major variables Thurow had in mind
when he concluded that "the distribution of human capital is an
important ingredient in the distribution of income, but it is not the
sole ingredient. The actual dispersion of income is much greater than
would be predicted by the distribution of human capital.u10
10Lester C. Thurow, Poverty and Discrimination (Washington:
The Brookings Institution, 1969), p. 109.
25
8
The Design of the Study
The dissertation proceeds in the following way to construct and
test a general model of wage determination. The major strands of a
general wage theory are discussed in Chapter II. Marginal productivity
theory, the institutional analysis, human capital theory, and social
stratification hypotheses are initially discussed. Each theory is
carefully weighed in order to glean material useful for developing a
testable wage determination model.
The general model is developed in Chapter III. A complete
theory of wage determination is first constructed which takes as its
premise a "deterministic" view of social relations. The distribution
of earnings is made a function of four exogenous variables: race,
sex, social class, and innate ability. Following this a specific
testable model is derived based on human capital, institutional, and
stratification hypotheses. The specific model is constructed so as
to hold human capital constant allowing the earning effect of industry
and occupation "crowding" to be measured. From this a reduced form
earnings generating function is developed. Chapter IV discusses the
econometric techniques used to measure the independent effects of
human capital and "crowding." Attention is paid to the potential
problem of multicollinearity and the statistical procedures used to
overcome them.
The statistical results follow in Chapter V. Regressions are
presented for five separate occupation groups which span the range of
all specific occupations in the United States. Individual regressions
i 26
9
are reported for each race-sex group as well as pooled race-sex
equations. In addition pooled regressions are presented which cover
the whole spectrum of occupations. The effect of individual human
capital, industry, stratification, and working conditions variables
is discussed.
An evaluation of the empirical results follows in Chapter VI.
Here the regressions are interpreted so as to parcel out the variance
in earnings due to human capital factors as a whole viz-a-viz labor
force stratification. Wage "ranges" are established for each
occupation group and each race-sex group based on a technique which
allows the human capital variables to be held constant while the
industry and stratification variables are permitted to vary together
according to empirically derived coefficients.
The final chapter is devoted to general conclusions and policy
implications. Emphasis is placed on the role of manpower policy in
the general antipoverty strategy. Some of the implications for
training programs and income maintenance schemes are explored.
Finally, there is some speculation as to the justification for the
present distribution of earned income, given the empirical results
found in this analysis.
There are two appendices in addition to the seven chapters.
Appendix A contains a description of the integrated macro-micro data
set with a discussion of its construction. Variables used in the
regression analysis are defined and the shortcomings in each is noted.
Appendix B contains the means and standard deviations for each
i Kwc.,44
10
regression as well as a complete set of zero-order correlation
matrices for all of the empirical results.
i 28
CHAPTER II
EXISTING THEORIES OF WAGE DETERMINATION
Individual prices reflect a near infinite set of past, present,
and even future events. Previous capital expenditures, the whole
galaxy of current prices of complementary and substitute products,
and expectations about future prices all impirge on the current
market value of each good. Consumer attitudes, changing tastes,
government subsidies, tariff policy, antitrust action and hundreds
of other factors interact to determine millions of prices.
Nevertheless, the key factors which determine the price of most final
and intermediate goods are well-known.
Yet economists have always been perplexed by one special case:
the price of labor.
Marshall, Pigou, Taussig, and other leading theorists weretroubled by the "peculiarities" of the labor market--the
fact that the worker sells himself with his services, that
his immediate financial n.,!ld may place him at a disadvantagein negotiating with employers, that he is influenced by
nonpecuniary motives, that he has limited knowledge of
alternative opportunities, and that there are numerous objective
barriers to free movement of labor.'
Numerous attempts have been made to fit the theory of wages
into a more general analysis of price. By assuming away a number of
1Lloyd Reynolds, The Structure of Labor Markets (New York:
Harper and Brothers, 1951), p. 2.
11
29
12
the "peculiarities" of the labor market, economists have treated
labor in the same manner as other productive inputs in the economy.
"The theory of the determination of wages in a free market is simply
a special case of the general theory of value," Hicks wrote. "Wages
are the price of labour; and thus, in the absence of control, they
are determined, like all prices, by supply and demand."2
The history of labor theory is rich in these abstract attempts,
but poor in empirical verification. The relative impact of supply
and demand forces on the wage rate remains, for the most part, a
mystery. Even Hicks admitted that "a long road has to be travelled"
before the concepts of wage theory "can be used in the explanation
of real events. H3 Barbara Wootton has responded that, "In practice
this road seems to have been not only long, but so exhausting that
few travellers have attempted it."4
Before setting out on this difficult road, it seems good
practice to review some of the theories developed in the past. Four
broad strands in the development of wage theory can be discerned:
(1) marginal productivity theory (2) institutional theory (3) human
capital theory, and (4) social stratification analysis. Most wage
theory fits into at least one of these categories.
2J.R. Hicks, The Theory of Wages (London: MacMillan, 1932), p. 1.
3Ibid., p. 10.
4Barbara Wootton, The Social Foundations of 1.7rage Policy (London:
Unwin University Books, 1955), p. 12.
30
13
What we are ultimately searching for is a theory that will
explain the empirically observed distribution of personal earnings.
Furthermore such a theory must be capable of describing and inter-
preting the relationship between the personal characteristics of the
individual worker and the wage he or she receives in the marketplace.
Of particular concern is the relationshipbetween the wages received
and a subset of personalcharacteristics which we shall call the
"endogenous productivitycharacteristics" of the individual.
By the term "endogenous productivitycharacteristics" we shall
refer to the innate and acquired physical and mental attributesof
the worker useful as inputs in the production process.This term is
synonymous with the term "human capital" when defined narrowly "as
an individual's productive skills, talents, and knowledge."5 This
new terminology is introduced because the term "human capital" has
been broadened in some recentliterature to include such factors as
race, sex, and the physicalattractiveness of the individual. While
these factors may be important in determining the distribution of
earnings, we find it useful to separate nut the personal character-
istics which would have no relationship to the distribution of earnings
in a "blind" economy--an economy in which theproductivity of an
individual was not related to color, sex, or physical beauty.6
Given
5Lester C. Thurow,Investment in Human Capital. (Belmont:
Wadsworth PublishingCo., 1970), p. 1.
6This, of course, should not be construed as to deny the
importance of these factors in the actual distribution of earnings.
31
14
this understanding the terms "endogenous productivity characteristics"
and "human capital" will be used interchangeably.
In the next section we shall introduce the term "endogenous
productivity" which relates to the potential output of an individual
given his or her endogenous productivity characteristics. Endogenous
productivity can be shown to be theoretically distinct from the more
common term, marginal productivity of labor.
Marginal Productivity Theory
Much of the debate over the distribution of earnings rests
on a more fundamental theoretical debate concerning the usefulness
of marginal productivity theory. Consequently the theory provides a
good starting point for any discussion of wage determination.
Orthodox or traditional wage theory rests on the fundamental
proposition that labor is paid its marginal product. "Workers are
paid according to how much they contribute to marginal increases in
output. If increasing the number of employed workers by one worker
would increase output by $5,000, workers should be paid $5,000."7
If
there are no additions to complementary inputs in the production
process, the wage of the "marginal" worker and all intramarginal workers
Physical attractiveness, for instance, may be the most important
personal attribute in some lines of work. Whether an individual
receives a particular job or not may be a function of physical
attractiveness and the actual market value of an individual in certain
occupations is a function of such factors as well. The distinction
between skill, for instance, and racial and sexual characteristics
is clear enough; physicalattractiveness falls into a grey area
somewhere in-between.
7Lester Thurow, Poverty and Discrimination, p. 26.
32
15
will be exactly equal to the full measure of any additional output.
This result follows according to marginal productivity theory
assuming no barriers to labor mobility, a homogeneous labor force,
and assuming that all employment yields homogeneous nonpecuniary
utility (or disutility). Where the product market is characterized
by reasonable competition and labor is freely mobile between employers,
the wage rate will equal the value of the marginal physical product.
wi= VMP1 = (MPP
i.11)
Where the product market is characterized by monopoly elements, the
equilibrium wage will equal the marginal revenue product.
wi= MRP = (MPP MR)
In either case an employer will not hire an additional worker if
revenue generated by that individual is less than the addition to his
wage bill. This assures the wage of labor will never be above its
marginal revenue product, at least as long as employers attempt to
maximize profits. The assumption of competition among employers for
labor services, on the other hand, assures the wage will not fall
below labor's marginal revenue product, and in the case of perfect
product markets, not below VMP.
Under conditions of monopsony in the labor market, labor is
paid less than its marginal revenue product.
wi< MRP
i
16
Monopsonistic employers face a rising supply curve rather than an
infinitely elastic supply of labor. Additional workers can only be
obtained by increasing the wage. In the absence of wage
"discrimination," the marginal cost of labor increases by more than
the wage bill paid to the marginal worker. The employer has to pay
the higher wage rate not only to the additional worker, but to all
of his workforce. Under these conditions, the marginal cost of labor
will lie above the wage rate, and equilibrium will therefore be
reached at a point where the marginal productivity of labor exceeds
the wage level.
The traditional analysis of wage determination has normally
been applied at either the level of the aggregate economy or at the
level of the firm; marginal productivity theory was not specifically
developed to explain the personal distribution of income. At the
level of economy, the supply of labor is assumed perfectly inelastic
or upward sloping. In this case the theory is useful as a theory of
aggregate wages. At the level of the individual firm, the supply of
labor is assumed perfectly elastic (with the exception of the
monopsony case) and the theory describes the level of employment in
each firm.
As long as there is perfect mobility of labor and labor is
homogeneous, there will exist a unique market clearing wage throughout
the economy. Each worker will receive exactly the same remunerationil
8This again assumes that all jobs have homogeneous nonpecuniary
utility. Where this assumption does not hold, "compensatory" wage
i '34
17
and each firm will hire just enough labor at this rate so as to keep
marginal revenue product (or the value of marginal product) from
falling below the market wage.
Classical marginal productivity theory paid little attention to
the characteristics of labor supply; labor was assumed homogeneous
throughout the economy or homogeneous within major categories or
broad occupations. The quality of labor was consequently accepted as
given. It was assumed that workers of a given quality could move
from one employer to another without interference. Relaxing the
labor homogeneity assumption, but retaining the assumption of perfect
mobility, yields a "modern" marginal productivity theory which is
theoretically capable of describing the personal distribution of
earnings among workers with different levels of human capital. As
Thurow has noted, "In an economy with perfect competition and in
equilibrium, the distribution of marginal products is identical with
the distribution of earned income."9
And,
the supply and demand for labor with differing skills
and knowledge would determine the marginal product of
each variety of labor. Individual earnings would equal
their marginal products, and the allocation of human
capital would determine the distribution of earnings.
differentials develop to account for differences in the nonpecuniary
advantages or disadvantages of particular jobs.
9Lester Thurow, Poverty.and Discrimination, p. 29.
10Ibid., p. 96.
35
18
Given free access to the labor and product markets consistent
with their human capital, workers will be distributed so as to
maximize the total value of output and in turn each worker's marginal
productivity and wage. If the equilibrium is disturbed in some way
(by the introduction of new technology, for instance), the labor force
will be reallocated so that once again the ordinal ranking of
endogenous productivity characteristics is consistent with maximized
output. From an efficiency standpoint, the wage structure under these
conditions will be optimal. From the viewpoint of "equity," the
personal distribution of earnings will be colinear with the distri
bution of human capital.
The colinear relationship between earnings and human capital
will occur whether the product market is characterized by perfect
competition or monopoly. However once monopsonistic elements are
introduced into the labor market, colinearity disappears. Labor of
given quality will receive less in the monopsonistic firm and the
differential will persist as long as mobility to other firms is
restricted. Where labor differs as to quality, the statistical
colinearity between human capital and earnings depends on which group
of workers is restricted to the monopsonized sector. In any case the
hypothesized link between endogenous productivity characteristics
and wage rates no longer holds.
The usefulness of the marginal productivity theory as a theory
of the distribution of earnings rests on the assumption of perfect
mobility of labor. To the extent that this assumption is violated in
1
frt0rib
10
19
the real world, the theory fails to adequately explain wage
determination. It fails for it specifies only one of the two critical
linkages between the distribution of earnings and the distribution of
human capital. The linkage between the wage rate and the marginal
revenue product of labor is well described by the theory. What is
not specified is the connection between MRP and the level of human
capital. This link relies on the nature of the labor market. It is
possible that every worker is paid his marginal revenue product at
the same time that his marginal revenue product bears no relationship
to his level of human capital. Given imperfect mobility of labor,
it is possible that:
(1) wi =MRPi
(2) MRP1 # f(Human Capital)
In this case a knowledge of the distribution of human capital would
be insufficient to describe the distribution of wage income.
At this point it is helpful to introduce a new term in order to
differentiate between the actual marginal revenue product of each
worker and the hypothetical marginal revenue product each worker would
receive if there were no barriers to labor mobility and the economy
were in equilibrium. This hypothetical marginal revenue product
shall be referred to as the "endogenous revenue product." The
endogenous revenue product of individual i (ERPi) is the marginal
revenue product individual i, possessing endogenous productivity
characteristics, C., would receive if he were to compete freely in the
20
labor market with all other workers with characteristics C.3 .
To clarify the distinction assume that all labor is of
homogeneous quality and there are two firms operating with identical
labor demand curves. Under the condition of perfect labor mobility,
all workers will earn a wage, wi, equal to the economy-wide marginal
revenue product, MRP*.11
Now introduce an arbitrary barrier to labor
mobility which results in three-fourths of the total labor force
being restricted to Firm B. (See Figure 2.1)
FIRM B
MRP*----
FIRM A
Employment in B F.B E* EA E* Employment in A
Figure 2.1 Marginal revenue products with
labor supply restrictions
11This result holds even if the two firms face different product
demand curves. In this case the total labor force will be allocatedso that wi = MRP
i* in each firm; shifts in the level of employment in
each firm will assure this result.
21
Under these circumstances the wage received by workers in Firm A will
be equal to MRPA while the wage received by workers restricted to
Firm B will be equal to MRPB. In this case the colinearity between
endogenous productivity characteristics and actual marginal revenue
product is nonexistent. The endogenous revenue product of each worker
is equal to MRP* while the actual marginal revenue products are
MRPA and MRPB.
For workers in Firm A:
MRPw.=A
> MRP* E ERPi
For workers in Firm B:
w = MRPB
< MRP* E ERPi
To repeat, marginal productivity theory describes the link between
wiand MRP
i'but fails to describe the relationship between MRP
iand
ERPi. Thus the traditional theory cannot be used as an earnings
distribution theory where labor immobility is extensive. To summarize:
Under the assumption of perfect labor mobility:
(I) wi= MRP
(2) MRPi = ERPi
(3) ERPi= f(C )
(4) wi = f' (C )1.
Under the assumption of imperfect labor mobility:
(1) wi= MRP1
(2) MRPi # ERPi
22
(3) ERP1 = f(C )
(4) w #(C.i)i
)
and the marginal productivity theory is no longer a theory of the
distribution of earnings.
Human Capital Theory
Traditional marginal productivity theory rests on two funda-
mental assumptions: (1) homogeneous labor supply, and (2) perfect
labor mobility. Institutional wage theory, to be discussed in the
next section, retains the first assumption but rejects the latter.
Human capital theory does the reverse. It extends the marginal
productivity theory to account for differences in labor quality, but
maintains that all workers of a given quality compete in the same
market. In assuming no barriers to labor mobility (for labor of the
same quality), the human capital theory is fully consistent with
traditional wage theory. All workers who have the same endogenous
productivity characteristics produce the same marginal revenue product
and earn the same wage.12
Thus
= MRPij
= ERPwigij
for all individuals, i, with human capital characteristics, j. For
the labor force as a whole, the distribution of earnings becomes a
function of the distribution of human capital.
12This holds, once again, except in the case of monopsony.
40
23
Renewed interest in the role of labor in the production process
began in the late 1950s. Edward Denison, in attempting to account
for the sources of economic growth in the United States in the
1929-1957 period, found that he could explain only 31 percent of the
increase in output if he were forced to assume that productivity of
labor did not change.13
T.W. Schultz, in a classic article on the
same subject, showed that gains in output over time could not be
solely attributed to increases in physical capital.14
The need arose
for a framework which stressed "human productivity" as a source of
economic growth. The idea of human capital was introduced into
economic analysis.
The new approach to the study of wages and employment did more
than merely add the dimension of labor quality to the traditional
productivity theory. It focused on the investment process by which
a given stock of human capital is accumulated. "Human capital models,"
according to Mincer, "single out individual investment behavior as a
basic factor in the heterogeneity of labor incomes."15
Empirically,
the human capital approach attempts to measure the individual and
13Edward F. Denison, The Sources of Economic Growth in the United
States and the Alternative;, Before Us (New York: Committee forEconomic Development, 1962), p. 266.
14T W. Schultz, "Investment in Human Capital," American Economic
Review, March 1961. Also in "Investment in Man: An Economist's View,"
Social Service Review, June 1959.
15Jacob Mincer, "The Distribution of Labor Incomes: A Survey,"
Journal of Economic Literature, March 1970, p. 6.
41
24
social rates of return from investments in formal education, on-the-
job training, vocational education, health care, additions to labor
market information, and migration. Assuming away labor market
imperfections, some research is even attempting to infer the
distribution of abilities from the distribution of earnings.16
The Basic Human Capital Model
Borrowing from Mincer and Becker, .e fundamental human capital
equation can be written as in equation 1.
(1) log Ex
= log E0+ rx
where Ex= earnings generated by investment x
r = market discount rate or the internal rate of
return on investment x
E0= earnings generated by other factors than
investment x
According to this simple formulation the earnings distribution is a
function of investments in education, training, and other individually
acquired human capital components (x) plus a function of E0which
include:: innate or natural ability and other factors. Given a market
determined r and assuming E0 constant, differences in earnings will be
directly related to differences in the amount of human capital
acquired by each individual.
16Mincer notes two articles in this genre: K. Bjerke, "Income
and Wage Distribution," Review of Income and Wealth, November 1970;
A.D. Roy, "The Distribution of Earnings and Individual Output,"
Economic Journal, September 1950.
42
25
In addition, according to Mincer, equation (1) can account for
imperfections in the product market, in the labor market, and in the
market for human capital investment funds. Unfortunately, however, it
does this in patchwork fashion. As Mincer notes
If the competitive assumptions are relaxed, internal rates
of return cannot be equated with the market rate of interest
and generally differ among individuals. Equation (1) can
remain serviceable, however, with r interpr4ted as a group
average internal rate of return on (investment x), while
individual differences in r and in log E0 are impounded in a
statistical residual.17 (emphasis added)
This solution to the market imperfection problem is useful as
long as the degree of imperfection is minor.18 But if a large part
of the variance in individual earnings is due to labor market
imperfections, the human capital model fails to specify the critical
variables and in fact may draw attention away from them. A similar
"error" term or shift coefficient could have been applied to the
marginal productivity model, but there too, the patchwork would have
been for the sake of "realisM" without yielding any analytic
18Two points are in order. In private correspondence,
Professor Harold Levinson has noted that even this formulation by
Mincer is not quite correct. The group average r may also be affected
by labor market imperfections. Restrictions in supply for a whole
occupation, for instance as in the case of the building trades or
the medical profession, would shift r itself rather than show up in
the residual. In this case, he argues, equation (1) would be assigning
some portion of Ex to investment x which is in fact related to
institutional factors rather than inve--ment.
The second point regards the relationship of this formulation
to the differentiation between "endogenous revenue product" and
"marginal revenue product" discussed in the section on the marginal
productivity theory. Clearly ERP is analogous'to Mincer's "group
average internal rate of return" while the "statistical residual"
43
26
improvement in the model.19
Reinterpreting equation (1) indicates that there are two factors
of importance in wage determination. One is the amount of investment
in human capital and the other is the rate of return earned on the
investment. Or as Thurow has formulated:
The value of human capital can be divided into price and
quantity components. Education and on-the-job experience
provide the principal means for increasing the quantity or
quality of an individual's capital. Migration, improvements
in information, and the elimination of market imperfections,
such as prejudice, are the chief instruments to raise the
price for existing capital. Although the price factor
would not exist in perfect markets where all were paid equal
amounts for the use of identical skills, in imperfect
markets it is an important element in valuing human
capita1.2°
The distinction is important and lies at the crux of much confusion
over the usefulness of the human capital needed. Depending on the
degree of imperfection in the labor markets, the effect of r on Ex
accounts for variance in MRP around ERP. In this case, r = ERP while
(MRP-ERP) = c, thy. statistical residual.
191n the recent literature there has been some attempt at
explicitly integrating market imperfections into human capital theory.
Much of this has focused on job search behavior. The job search is
viewed as another form of investment in human capital where the costs
of the search, including opportunity costs, must be weighed against
potential discounted future earnings. While this tends to account
for the problem of "imperfect" markets due to information cost, it
fails to solve the larger problem of imperfect mobility due to market
discrimination. See Charles C. Holt and Martin H. David, "The Concept
of Job Vacancies in a Dynamic Theory of the Labor Market" (Madison:
Social Systems Research Institute, University of Wisconsin, 1965) and
Dale T. Mortensen, "A Theory of Wage and Employment Dynamics"; and
Donald A. Nichols, "Market Clearing for Heterogeneous Capital Goods,"
in Edmund S. Phelps, Microeconomic Foundations of Employment and
Inflation Theory (New York: Norton, 1970).
20Lester C. Thurow, Poverty and Discrimination, op. cit., p. 69.
44
27
may outweigh the effect of 1. How much each individual invests in
human capital may not be as important as the rate of return he or
she receives on that investment. For a given population the distri-
bution of r's may be such as to reduce significantly the covariance
between the distribution of x's and Ex's. In this case,
The concept of human capital loses its economic meaning.It no longer reflects productive capacities, and it nolonger can be viewed in the same light as physical capital.In a fundamental sense the problems of determining individualincomes cease to be economic and become sociological orinstitutional.21 (emphasis added)
Empirical Studies of the Human CapitalEarnings Function
The development of the human capital function was followed by
a steady flow of empirical studies aimed at quantifying the deter-
minants of earnings. Many of the earlier studies attempted to measure
the private and social returns to education by estimating the
discounted present value of investment in formal schooling.22
Other
21 Lester Thurow, Investment in Human Capital, op. cit., p. 18.
22Some of the more important studies in this area include:
Gary Becker, Human Capital (New York: National Bureau of EconomicResearch, 1964); W. Lee Hansen, "Total and Private Rates of Return on
Investment in Schooling," Journal of Political Economy, April 1963;
Orley Ashenfelter and J.D. Mooney, "Some Evidence on the PrivateReturns to Graduate Education," Southern Economic Journal, January1969; M. Blaug, "The Private and Social Returns to Investment inEducation: Some Results for Great Britain," Journal of Human Resources,
Spring 1967: A.B. Carroll and L.A. Ihnen, "Costs and Returns for TwoYears of Postsecondary Technical Schooling," Journal of PoliticalEconomy, December 1967; W. Lee Hansen, Burton Weisbrod, and W.J.Scanlon, "Schooling and Earnings of Low Achievers," American Economic
Review, June 1970; E.F. Renshaw, "Estimating the Returns to Education,"
Review of Economics and Statistics, August 1960; E.A.G. Robinson and
45
28
studies more explicitly analyze the distribution of income and
earnings with human capital factors as the independent variables.
Using diverse earnings functions, a number of studies have
attributed a large part of the explained variance in incomes to
differences in education. Morgan, David, Cohen, and Brazer, using
multiple classification analysis on national survey data, found the
most important factor determining hourly earnings of household heads
is an education-age interaction term.23
The beta Loefficient on the
education-age term was .234, highest among the fourteen variables in
their analysis including sex, occupation, race, geographic mobility
and several general demographic factors.24
Using the 1/1000 1960
Census sample, Giora Hanoch finds a relatively high "marginal product"
for education, although education appears in his formulation to be
subject to diminishing returns.25 The result is similar to Weisbrod's
findings for private rates of return on different levels of schooling.
J.E. Vaizey (eds.), The Economics of Education (Londol: St. Martins,
1966); Gerald Rose, Differential Returns to Investments in Human
Capital in the Academic Labor Market, University of California
(Unpublished Ph.D. dissertation, 1969).
23James N. Morgan, Martin H. David, Wilbur J. Cohen, and Harvey
E. Brazer, Income and Welfare in the United States (New York:
McGraw-Hill, 1962).
24Ibid., p. 48.
25See Giora Hanoch, "Personal Earnings and Investment in Schooling,"
Ph.D. dissertation, University of Chicago, 1965; also, "An Economic
Analysis of Earnings and Schooling," Journal of Human Resources, Winter,
1967.
i 46
29
Lowell Gallaway,26
H.S. Houthakker,27
Elizabeth Waldman,28
and
Herman Miller29
have all attached a great significance to formal
education in explaining the distribution of earnings.
More recently, however, a good deal )f research has called
into question the great importance of formal education in earnings
functions. This is especially true of studies directed at explaining
the income differences of whites and nonwhites. Using a 77 cell
education-occupation matrix, Bluestone et al. find that a maximum of
30.3 percent of the income differential between full-time, full-year
employed white and black men can be explained by the quantity of
formal schooling.30 Two-thirds or more of the differential is due
to occupational discrimination (education statistically held constant),
discrimination in industrial attachment, and human capital factors
not included in formal schooling. Only 2.8 percent of the total
differential between full-time employed white men and white women can
26Lowell Gallaway, "The Negro and Poverty," Journal of Business,January 1967, pp. 27-35.
27H.S. Houthakker, "Education and Income," Review of Economics
and Statistics, February 1959.
28Elizabeth Waldman, "Educational Attainment of Workers,"
Monthly Labor Review, February 1969.
29Herman P. Miller, "Annual and Lifetime Income in Relation to
Education," American Economic Review, December 1960.
30The calculations were made from data obtained in the 1967Survey of Economic Opportunity. These specific calculations can beobtained from the authors. Similar results in a more disaggregated
model can be found in Barry Bluestone, Mary H. Stevenson, and WilliamM. Murphy, Low Wages and the Working Poor (Ann Arbor: Institute of
r 47
30
be explained by formal schooling; 14 percent of the differential
between black women and white men. Donald Katzner found a similar
result for the white/nonwhite earnings differential.31
Michelson,
criticizing earlier studies for failing to account correctly for the
interaction between education and occupation, finds the effect of
schooling on earnings to be even smaller.32
Using a larger matrix,
Michelson shows "that only 16 percent of the earnings differential
between whites and nonwhites would have been corrected by equal
schooling categories, employing current (1959) earnings per year of
school for each racial group."33
In response to the evidence that formal schooling explains only
a fraction of the earnings differential, especially among race-sex
groups, additional human capital variables have been added to the
earnings function. A catalog of these variables compiled by Hansen,
Weisbrod, and Scanlon includes: (1) physical condition, including
general state of health and specific disabilities; (2) mental
capability, reflecting inherited potential; (3) learning and
experience, determined not only by the quantity and quality of formal
Labor and Industrial Relations, Research Division, University of
Michigan-Wayne State University, 1973).
31Donald A. Katzner, Theory and Cost of Racial Discrimination,
Ph.D. dissertation, University of Pennsylvania, 1966.
32Stephan Michelson, Incomes of Racial Minorities (Washington:
The Brookings Institution, 1968) unpublished manuscript.
33Ibid., pp. 2-35.
I 48
31
education, but by specific job training and job experience; (4)
psychological characteristics, among them motivation and ability to
communicate and cooperate in work situations; and (5) family
environment, reflecting informal learning, socialization, and
n contacts."34
The study by Morgan, et al. attempted to explain hourly earnings
using proxies for many of these factors. While "supervisory
responsibility," "attitude toward hard work and need-achievement
score," "interviewers' assessment of ability to communicate," and
"physical condition" were statistically significant, each of these
factors explained only a minute fraction of the variance in wage
rates after controlling for other variables.35
Altogether their
fourteen variables including education and age, sex, occupation, race,
and geographical location (in addition to the preceding variables)
accounted for 34 percent of the variance in wage rates. Two-thirds
remained unexplained.
Other researchers have continued to add new variables to the
basic human capital model in an attempt to explain the variance in
earnings. Chief among these are the quality of education, work
experience, and on-the-job training. Johnson and Stafford used
educational expenditure per pupil as a proxy for "quality" and found
34W. Lee Hansen, Burton A. Weisbrod and William J. Scanlon,
"Determinants of Earnings: Does Schooling Really Count?" Discussion
Paper, Institute for Research on Poverty, University of Wisconsin,
August 1968 (Preliminary Draft).
35Morgan et al., Income and Weifare in the United States,
Chapter 5.
49
32
"that there are high but diminishing marginal returns to investment
in school quality.06 Thurow has used years of experience in the
labor force as a proxy for "experience" or on-the-job training.37 Rees
and Shultz find "seniority" with the present employer to be the most
significant variable in explaining the wages of workers in their
Chicago labor market study.38 The use of "age" in other studies is
intended to act partially as a proxy for experience. In each case,
age, experience, general on-the-job training, or seniority has been
found significant. Yet with few exceptions even the most complete
human capital equation seldom explains more than 35 percent of the
variance in income and usually the explanatory power of such models,
measured in terms of R2
, is much lower.
Of course, a relatively high R2 only indicates correlation; it
indicates nothing about the causal nature of the relationship or even
the direction of causation. This is especially important for human
capital functions. In the case of experience, for instance, the
36George E. Johnson and Frank P. Stafford, "Social Returns to
Quantity and Quality of Schooling," Department of Economics, University
of Michigan, 1970, unpublished manuscript, pp. 17-18. Other works on
school quality as an input in the human capital equation include:
Finis Welch, "Measurement of the Quality of Schooling," American
Economic Review, May 1966; and James Morgan and Ismail Sirageldin, "A
Note on the Returns to Quality of Schooling," Journal of Political
Economy, September-October 1968.
37Lester C. Thurow, "The Occupational Distribution and Returns
to Education and Experience for Whites and Negroes," Federal Programs
for the Development of Human Resources, Joint Economic Committee
(Washington, D.C., 1968), pp. 267-84.
38Albert Rees and George B. Shultz, Workers and Wages in an Urban
Labor Market (Chicago: University of Chicago Press, 1970), pp. 84-85.
50
33
problem is severe. There are a number of interpretations of the
relationship between experience and earnings, all of which are con-
sistent with the data, but which point to extremely different hypotheses
about the determinants of income. One is that experience adds
directly to a worker's endogenous revenue product and therefore is a
legitimate human capital variable. Another, however, views experience
or seniority as reflecting nothing more than institutionalized pay
increments based on length of service and set out in collective
bargaining agreements or offered by employers to maintain morale. In
this case, "experience" may be totally unrelated to changes in an
individual's human capital.
Of even greater damage to the human capital interpretation of
"experience" is the possibility that the causal link between earnings
and experience may actually be reversed. Higher wages may cause
longer experience. Workers in "high wage" industries may have lower
turnover rates and therefore longer on-the-job experience because there
is little room for improving earnings by moving to new employment.
Workers in "low wage" industries or firms may feel less attachment
to their present employer and search often for new jobs. In this case
seniority may be low, but possibly the result of low wages rather than
the reverse.39 In addition, the existence of "internal labor markets"
39Doeringer has shown that this occurs often in ghetto labor
markets. In a study of a Boston manpower program, he found that job
tenure was directly related to wage level after controlling for other
factors. See Peter B. Doeringer, "Manpower Programs for Ghetto Labor
Markets," Proceedings of the 21st. Annual Winter Meeting of the
Industrial Relations Research Association, pp. 257-267.
51
34
produces a situation where training and experience become a function
of being hired.40
Similar ?roblems of interpretation exist with other human
capital variables as well. The relationship between years of formal
schooling and level of human capital is by no means clear. Bowles,
for instance, has begun work on educational production functions to
determine what inputs from formal schooling contribute to "pro-
ductivity."41
Of special concern is whether schooling actually con-
tributes to the endogenous revenue product of the individual or whether
the empirically derived relationship between education and earnings
merely reflects the use of formal schooling as a "credential" in the
employment screening process. Other human capital variables suffer
the same problems of interpretation. General and specific job
training, IQ scores, health and disability measures, and factors
contributing to geographical mobility all appear to contribute in one
way or another to the "explanation" of earnings. But the precise
connection between independent and dependent variables remains fuzzy.
The more important problem with the human capital theory remains,
however, even if the problem of causal relationships is set aside.
Like its predecessor in productivity theory, human capital models fail
40For the best discussion of internal labor markets, see Peter B.
Doeringer and Michael J. Piore, Internal Labor Markets and ManpowerAnalysis (Lexington: D.C. Heath, 1971).
41 See Samuel Bowles, "Towards an Educational Production Function,"in W. Lee Hansen (ed.), Education, income and Human Capital (New York:National Bureau of Economic Research, 1970).
52
35
as a theory of personal earnings where labor markets are imperfect.
The use of "race" and "sex" variables as human capital components or
"quality" variables clearly improves the fit of so-called human
capital functions. But it should be equally clear that such variables
are quite distinct from what we have called "endogenous productivity
characteristics." Where racial and sexual discrimination exist in the
labor market, or where mobility barriers are established through
monopsonistic power or trade union practice, the distribution of wage
income and the distribution of endogenous revenue products need not
be covariant. In this case explicit attention must be paid to the
institutional factors in the economy which impinge on the distribution
of earnings. If mobility barriers are important in the economy, the
ad hoc addition of new "human capital" variables may boost R2a bit,
but will add little to a meaningful explanation of wage determination.
The Institutional Approach
Whereas the marginal productivity theory and human capital
theory for the most part ignore the existence of barriers to labor
mobility, the institutional approach to wage theory begins with the
basic position that market imperfections are sufficiently widespread
to cause wage rates to deviate significantly from their free market
equilibrium levels. Thus, according to institutional theory, an
individual's actual marginal revenue product can diverge significantly
from his endogenous revenue product.
C3
36
The institutional approach, developed in the late 1930s, came
in response to rising unionism, a growing awareness of monopoly
elements in the economy, and increased government intervention in the
marketplace. Rather than an immediate concern with the determinants
of individual earnings, institutional theory has attempted to untangle
the various factors which impinge on interindustry and interregional
wage distributions. Where the marginal productivity theory focused
on absolute wage levels, the institutionalists have been more con-
cerned -with relative wages (and changes in relative wages) for similar
work. Assuming the "quality" of labor homogeneous within a given
occupational range, the institutional approach investigates the impact
of such factors as unionization and the effect of "ability to pay" on
relative wages. Lacking the theoretical rigor of other approaches
to wage determination, the institutional approach compensates with
vigorous empirical investigation.
Balkanized Labor Markets
Adam Smith attempted to explain wage differentials by noting
"compensating" differences in job content. In contrast, J.S. Mill
argued as early as 1847 that wage differentials are due to the absence
of competition in the market for labor.42 Stressing the existence of
42J.S. Mill, The Principles of Political Economy with Some of
Their Applications to Social Philosophy in Vol. II, Collected Works
of John Stuart Mill (Toronto: University of Toronto Press, 1965).
According to Mitchell, the first reference to "noncompeting" groups
in labor is found in Mill's lecture notes. See Wesley C. Mitchell,
Types of Economic Theory, Vol. I (New York: Augustus M. Kelley, 1967),
p. 562n.
54
37
barriers to occupational entry, Mill pictured the labor market as
deeply fragmented with individual workers falling into specifically
defined markets. Little intermarket mobility could be expected between
"noncompeting" groups of labor. Barriers to mobility between
occupations, in Mil: iew, were due to the social class structure
of who we would today call human capital investment behavior.
The theme of barriers to labor mobility is implicit in the
institutional anal} 's. Pat in place of strict family occupational
lines characteristic of a preindustrial era, the modern labor market
is "bulkanizea" or segmented into many sub-markets by institutionally
developed rules, both formal and informal.43
Entrance into each sub-
market, and movement within its internal market channels are often
strictly defined. Tha degree of unfettered choice within the overall
labor market is consequently diminished. Those who gain access to
restricted markets presumably gain woges higher than they normally
would in the face of perfect competition.
Restrict,,n of employment in any one sub-narket or firm occurs
in one of two ways (or both). In markets controlled by strong
employee organizations (e.g. building trades, the medical profession)
the actual supply of labor may be restricted to some given level.
The intersection between the market demand curve and the "institu-
tionalized" st ply curve yields the sub-market wage. In other markets,-...
43See Clark Kerr, "The Bulkanization of Labor Markets," in
E. Wight Bakke, Labor Mobility and Economic Opportunity (Cambridge:
The New 1,!chnology Press and John Wiley, 1954).
55
supply is not explicitly regulated, but the wage level is. In this
case the amount of labor employed in the sub-market is a function of
sub - market demand and the institutionally set wage. In both cases it
is presumed that the resulting wage exceeds the wage that would exist
in perfect competition.
The existence of strong unions on the labor supply side is an
important element in the institutional framewot.c. Yet other factors
which contribute to the balkanization of labor markets--racial and
sexual discrimination, barriers to geographical mobility, "lock-in"
effects of seniority, civil service channels, etc.--contribute to the
institutionalist argument that wage rates may reflect other factors
beside the'endogenous productivity characteristics of labor.
Balkanized labor markets, however, are only part of the insti-
tutionalist approach to wage theory. On the demand side, institu-
tionalists argue that firms do not profit maximize, that the marginal
conditions needed to maximize profits are not and cannot be known with
accuracy, and that firms have other goals with respect to their
workforces beside maximizing output at minimum labor cost.44
Instead
of reflecting the lowest wage possible in every instance, firms with
an "ability to pay" will often offer higher wages that necessary to
secure the quantity of a given quality of labor it desires. In the
44For the best summary of the institutionalist attack on
marginal productivity theory and for one of the strongest rejoinders,
see the Richard Leste.-Fritz Machlup "debate." This appears in three
issues of the American Economic Review, March 1946, September 1946,
and March 1947.
56
39
words of a leading institutionalist, "The major factor [in wage
determination) is differences in companies' wage - paying ability, plus
in some cases the presence of a union which forces a company closer
to the limit of its ability to pay."45
Thiq stress on "ability-to-pay"
has led many empirical studies to focu, on those factors related to
the well-being of a firm or industry: ,centration, profits,
capital-intensity, and productive efficiency. Unionization takes on
the role of a political power variable in addition to its role in
restricting labor supply.
An Institutional Model of WageDetermination
Pulling together the separate strands of institutional wage
theory allows the development of a unified institutional theory of wage
determination.46 Labor supply is assumed homogeneous in quality but
45Lloyd Reynolds, The Structure of Labor Markets (New York:
Harper and Brothers, 1951), p. 189.
46Some of the more important research which went into the insti-
tutional synthesis included: John Dunlop, Wage Determination UnderTrade Unions (New York: John Wiley, 1944); Arthur M. Ross, Trade Union
Wage Policy (U. of California, 1948); Harold M. Levinson, Unionism,
Wage Trends, and Income Distribution z_ 1914-1947 (Ann Arbor: University
of Michigan Press, 1951); Lloyd G. Reynolds, The Structure of Labor
Markets (New York: Harper and Brothers, 1951); Sumner Slichter, "Notes
on the Structure of Wages," Review of Economics and Statistics,
February 1950; Lloyd G. Reynolds and Cynthia H. Taft, The Evolution of
Wage Structure (New Haven: Yale University Press, 1956); William Bowen,Wage Behavior in the Postwar Period (Princeton: Princeton University,
Industrial Relations Section, 1960); Harold M. Levinson, PostwarMovement of Prices and Wages in Manufacturing Industries, JointEconomics Committee, 86th Congress, 2nd Session, Study Paper No. 21,
1960; Albert Rees, "Union Wage Gains and Enterprise Monopoly," Essayson Industrial Relations Research (Ann Arbor: University of Michigan,
57
40
balkanized by factors other than those related to endogenous
productivity characteristics. Product markets are differentiated by
entry barriers, either accounted for by scale requirements or spatial
area limitations.47 And unions are responsible for Picher directly
limiting labor supply or using their bargaining power to raise wage
levels above the competitive norm.
Wages are then determined through a complex interaction of
economic constraints, political decisions which affect the strength of
unionism, and finally the relative bargaining power of labor and
management.48
INDUSTRY WAGE DIFFERENTIALS
ECONOMIC FACTORS4t----- ------>BARGAINING POWER FACTORS41
POLITICAL FACTORS
At any given pcint in time, the general level of physical productivity
and market demand conditions place an upper limit on the final wage
Institute of Labor and Industrial Relations, 1961); H. Gregg Lewis,
Unionism and Relative Wages in the United States (Chicago: University
of Chicago Press, 1963; Harold M. Levinson, "Unionism, Concentration,
and Wage Changes: Toward a Unified Theory," Industrial and Labor
Relations Review, January 1967; Harold M. Levinson, Determining Forces
in Collective Wage Bargaining (New York: John Wiley, 1968).
47The argument about spatial limitations of the physical area
of a labor market is developed in Harold M. Levinson, "Unionism,
Concentration, and Wage Changes: Toward A Unified Theory," op. cit.
48This model is most thoroughly discussed in Harold M. Levinson,
Determining Forces in Collective Wage Bargaining, op. cit.
58
41
bargain. The hypothetical competitive labor supply curve places
the lower limit on the -sage bargain. The final wage settlement will
then lie within this range and be determined by the relative
bargaining power of employer and employee representatives and/or the
nature of the preference function of management where union strength
is either weak or nonexistent.
The institutional synthesis can be depicted as in Figure 2.2.
In the short run, firms are faced by a wage range, WW - Wm, where W
is the reserve price of labor (of a given quality) in the absence of
any restriction of labor supply. At a wage below Wc, firms will find
no one willing to work. Above Wm, firms will cease all production
because WM> MRP at all levels of output. Through collective
Employment
Figure 2.2 The institutionalized wage bargain
59
. .
42
bargaining, the labor supply curve will be raised vertically so that
the height of Su above We ,:epends on the relative bargaining strength
of labor and management. Increases in the marginal physical product
of labor (through increases in the capital/labor ratio or techno-
logical advances) or increases in marginal revenue (due to changes in
market demand conditions) will shift the MRP curve up and the wage
band will expand to WC M
- W' The final wage W* thus depends on economic
variables (marginal physical product and marginal revenue) and
bargaining power factors. Behind the scenes, the government plays a
political role in modifying the relative strength of labor and
management.49
Even with homogeneous labor supply, wage rates can differ
between firms or industries depending on the relative height and slope
of the marginal revenue product curves and the relative strength of
labor and management. The institutional model consequently predicts
the following results:
(1) Individuals barred from protected industries will earn lower
wages than individuals in other industries even where endogenous
revenue products are equal.
49In the long run W* is indeterminate without knowledge of the
elasticity of substitution between capital and labor. Autonomous
increases in the price of labor may drive firms to raise the capital-
intensity of their production processes. In this case, the Su curve(s)
and demand curves will no longer be independent. The precise wage
outcome requires information about the elasticity of substitution
between capital and labor and the wage preference function of union
leadership.
1 60
43
(2) Wages will reflect industry characteristics as well as the
endogenous productivity characteristics of the workforce.
(3) In particular wages of similarly qualified individuals will
be higher for those who gain access to unionized industries and
industries where competitive pressures are minimized through monopoly
or spatial limitations to market entry.
Empirical Verification of the
Institutional Approach
With few .ceptions data on manufacturing industries have been
used to test the institutional predictions. The usual dependent
variable is change in average straight-time hourly earnings over
time.50 The critical independent variables have been unionization
and concentration while some attention has focused on profits, change
in sales, value-added, and capital-labor ratios.51
50Changes or increasesin average wage rates rather than
absolute wage levels have been used in institutional analyses in order
to circumvent the problems caused by differences in the "quality" of
the labor force in different industries. Presumably while labor
quality may vary from industry to industry, changes in the average
endogenous productivity of an industry's workforce come only slowly.
Thus there should be little relationship between changes in wage rates
and changes in the quality of the labor force. Any significant corre-
lation between industry factors and changes in wage rates should
consequently be free of hidden correl:'tion with human capital variables.
Unfortunately this does not completely solve the problem, however.
51 For early work using some of these industry factors see:
Sumner Slichter, op. cit.; John T. Dunlop, "Productivity and the Wage
Structure," in Income, Employment and Public Policy, Essays in Honor
of Alvin H. Hansen (New York: Norton, 1948); Joseph Garbarino, "A
Theory of Inter-Industry Wage Structure Variation," Quarterly Journal
of Economics, May 1950.
61
44
In an analysis of data for the 1923-40 period, Garbarino found
a zero-order correlation of .7702 between the rate of increase in
earnings and the degree to which output is concentrated in a few
firms.52 In a similar study of 50 industries covering the period
1933-1946 Ross and Goldner found a very strong correlation between
concentration and changes in average straight-time hourly earnings.53
Significant positive relationships were also found with unionization
and growth in employment, although Ross and Goldner admitted that they
could not disentangle the independent effects of concentration, employ-
ment growth, and unionization. They concluded that concentration and
growth provide a "permissive" economic environment within which unions
can appropriate a portion of monopoly profits. In a later multiple
regression analysis, Bowen confirmed the difficulty of untangling the
institutional variables, but found that wages rose more rapidly in
industries with rising employment, higher profits, higher concentration
ratios and stronger unionization.54 Finally Levinson confirmed the
importance of profits, concentration, and unionization.55 He found a
strong relationship between earnings, lagged profits, and 1954 concen-
tration ratios, but found no general relationship between union
52Joseph Garbarino, op. cit., p. 302.
53Arthur M. Ross and William Goldner, "Forces Affecting the
Interindustry Wage Structure," Quarterly Journal of Economics, May 1950.
54William Bowen, op. cit.
55Harold Levinson, Post-War Movement of Prices and Wages in
Manufacturing Industries, op. cit.
62
4c5
strength and wage changes. The importance of union strength per se,
he argued, does not show up statistically, but nevertheless exists
through pattern and demonstration effects.
H. Gregg Lewis has derived numerous estimates of the effect of
unionization in different sectors of the economy f different
historical periods.56 Most of these estimates use data compiled by
a number of his students. Using data developed by Sobotka, Lewis
estimated that in 1939 unionization increased wages of common laborers
in the building construction industry by approximately 5 percent.
Skilled craftsmen, however, members of much stronger unions, were able
to raise their relative wage 25 percent through organizing.57
In
bituminous coal, the effect of unionism ranged from a high of over
120 percent in the early 1920s to zero at the end of World War II.58
The effect of unionization on relative wages in other industries
included 10-18 percent in rubber tire manufacturing (1948), 7 percent
in wooden furniture manufacture (1950), 19 percent for barbers (1954),
and 6-10 percent for hotel employees (1948).
The weighted average effect of unionism on wages in the 12
industry studies Lewis reviewed is 18 percent. In cross-industry
56H. Gregg Lewis, op. cit.
57The construction industry estimates derived by Lewis are
based on Stephen Sobotka, "The Influence of Unions on Wages and Earnings
of Labor in the Construction Industry," Ph.D. dissertation, University
of Chicago, 195'7..
58Based on Rush V. Greenslade, "The Economic Effects of
Collective Bargaining in Bituminous Coal Mining," Ph.D. dissertation,
University of Chicago, 1952.
63
46
global studies of interindustry wage variation, Levinson,
Garbarino, and Goldner found an effect of similar magnitude while
Ross and Ross and Goldner found a somewhat smaller effect (although
biased by the 1945-46 period). Levinson's estimate was in the
neighborhood of 17 percent while Garbarino and Goldner were closer to
15.59
Problems with The Institutional Approach
In relaxing the assumption of labor mobility inherent in tra-
ditional wage theory, the institutional approach should be an important
addition to an understanding of wage determination. Unfortunately,
however, there are a number of problems in institutional analysis which
detract from its general usefulness.
One of these is the difficulty in specifying and measuring such
abstract factors as "ability-to-pay" and "bargaining power." The use
of profit rates and concentration clearly do not serve as strong
proxies for either of these and the proi-ortion of employees covered by
collective bargaining agreements--the normal measure of "unionization"- -
certainly leaves something to be desired as a measure of restricted
59These estimates were derived by Lewis by correcting the
earlier estimates in the original studies in order to make them con-
sistent. For the detail on these studies, see, Harold M. Levinson,"Unionism, Wage Trends, and Income Distribution, 1914-1917," MichiganBusiness Studies (Ann Arbor:Bureau of Business Research, GraduateSchool of Business, University of Michigan, 1951); Joseph W. Garbarino,op. cit.; William Goldner, "Labor Market Factors and Skill
Differentials in Wage Rates," Industrial Relations Research Association,Proceedings of the Tenth Annual Meeting, 1958, pp. 207-16.
i 64
47
labor supply or bargaining power. Consequently even if industry
characteristics seem able to account statistically for differences
in wage levels or changes in wage levels, it is difficult to relate
the results of the reduced form equation to a specific theory of wage
determination.
A second problem, at least for present purposes, is that insti-
tutional wage theory does not account for the stratification of certain
workers into certa n industries. It generally assumes imperfect
mobility without specifying the parameters of stratification. While
there is some theory as to why certain industries become concentrated
or unionized, there is practically no hypothesis as to which workers
will be employed in the more unionized industries and which in more
competitive sectors of the economy. This problem, of course, stems
from the fact that institutional theory never was intended as an
explicit theory of the personal earnings distribution.
There is an additional problem, however, which is potentially
more serious than either of these. It is possible that the insti-
tutional model is partially misspecified.60
The correlation between
earnings and institutional variables may be partly spurious hiding a
60The term "misspecified" in this context does not simply refer
to the fact that one or more variables in the model may be specifiedin the wrong mathematical form (e.g. linear rather than log normal).Rather misspecification refers to the possibility that there doesnot exist a real causal relationship between the endogenous variableand the several exogenous factors. In this case the significantcorrelation found between variables is spurious. A correctly
specified model would be one in which the causal relationship betweenvariables is theoretically sound.
65
48
correlation between earnings and endogenous productivity character-
istics. Misspecification may result because of the sequential
ordering of the acquisition of human capital, the determination of
occupation and industry, and the receipt of earnings. If occupational
and industrial attachment is a function of the level of human capital,
some of the variance in earnings normally attributed to industry
characteristics may in fact be due to differences in human capital
factors correlated with industry variables, but unspecified in insti-
tutional models. More formally, to the extent that (1) this corre-
lation exists and (2) the acquisition of human capital is causally
prior to job placement, significant coefficients on industry variables
are spurious and due to errors in the specification of the institutional
model. In the extreme case, the true relationship is between human
capital and earnings and the institutional model vanishes.
The use of data on wage ch;:ages rather than wage levels does not
completely eliminate this problem. The extent of unionization or
concentration across industries may be perfectly colinear with the
industry distribution of human capital. In this case both industry
and human capital variables would equally describe interindustry wage
changes and there would be no way a priori to determine which is the
true relationship. It may be true that larger wage increases as well
as higher wage levels are accorded higher skilled workers.
A fair test of the effect of institutional variables thus
requires a model which explicitly accounts for differences in human
capital and furthermore specifies the relationship between industry
49
factors and endogenous productivity characteristics. Such a model
would improve our understanding of interindustry wage differentials.
To go even further and develop a complete and coherent theory of the
distribution of personal earnings requires additional information
on the social stratification process in labor markets.
Social Stratification Analysis
Sociological literature since the time of Marx is replete with
attempts to understand the development and structure of social
stratification. Marx's theory of class conflict placed stratification
at the root of all social change.61
He viewed man's relation to the
means of production as the primary determinant of economic structure
and class differentiation. Emile Durkheim similarly placed great
emphasis on the division of labor.62
Increasing social density, he
argued, led to increasing occupational differentiation, lessened
social consensus, and altered the nature of social solidarity.
Other sociologists have studied the nature of status and prestige.
Weber investigated the relationship between social and economic
orders, stressing the importance of status as differentiated from
economic standing.63
Others have attempted to distinguish t,.t
61For an excellent review of Marx's theory of class differ-
entiation, see Reinhard Bendix and Seymour M. Lipset (eds.), Class,Status and Power, "Karl Marx's Theory of Classes" (Glencoe, FreePress, 1953), pp. 6-12.
6 2Emile Durkheim, On the Division of Labor in Society (New York:
Macmillan, 1933).
63Max Weber, "Class, Status, and Party," in Max Weber, Essays in
Sociology (New York: Oxford University Pref.s, 1946).
50
differences between class, status, and prestige. Within this broad
thrust is a more specific inquiry into the roots of social and
occupational mobility.64
For present purposes, a narrower perspective on social strati-
fication is ne:essary. A general theory of the personal earnings
distribution requires that the mechanisms of occupation, industry,
and wage stratification be clearly described. While this cannot be
done here in detail, a brief taxonomy of economic stratification is
helpful.
Two distinct mechanisms of differentiation can be identified.
One follows from human capital theory and the other from the insti-
tutional analysis. The former relates to differential access to human
capital; the latter to differences in the rates of return on a given
set of endogenous productivity characteristics. Both are related to
differences in race, sex, and social class.65
(1) Access 'to human capital. Investment in human capital can
vary between individuals for numerous reasons. Differences in time
preference, for one, can make a large difference in how much and when
64See, Pitirim A. Sorokin, Social Mobility (New York: Harper,
1927). For a review of social mobility studies, see S.M. Miller,"Comparative Social Mobility," Current Sociology, 1960.
65By the term "social class" we shall refer to a group of
individuals who generally possess common economic and social character-istics. The key determinants of social class by this definition includeincome, wealth, consumption, and social status. Social status isnormally conferred through one's occupation. The intergenerationaltransfer of "social class," for empirical purposes, is measured byoccupational standing and/or income.
68
51
individuals invest in themselves. Differences in ability may affect
investment rates as well. Individuals with greater innate ability,
for instance, may tend to remain in school longer and invest in more
training especially if training and ability interact to produce
extraordinary returns. Differences in income-leisure preference may
also tend to differentiate human capital investment. In each of these
cases, differential amounts of investment may be said to be
"voluntary."66
There are other cases, however, where differential investment is
involuntary and reflects social stratification. Because human capital
investment funds are not a "free" good and the market for investment
funds is imperfect, social class, race, and sex can enter into the
determination of each individual's stock and structure of acquired
human capital. The level of private investment often reflects the
level of personal income while the level of social investment (e.g.
thrimck public schools) is dependent on the social stratification
among legal jurisdictions.
Given imperfect human capital markets, wages can differ con-
siderably among individuals even if innate abilities and personal
preferences are identical, product and labor markets are perfect, and
66Extreme caution must be exercised in the use of the term"vcluntary." An individual's time preference, for instance, probablydepends on a whole set of factors, many of which he cannot control.Family attitudes, the social millieu, and economic conditions mayall p.ay a role in determining an individual's time preference andincome-leisure trade-off. "Culture" obviously influences a person'smotivation and may well be associated with social class.
69
52
rates of return are equivalent everywhere for identical levels of
human capital. Under these conditions, if acces. to human capital
is influenced by such characteristics as t-ce, sex, and social class,
wage differentials will reflect these factors.
(II) Differential rates of return. Differential amounts of
investment are sufficient to produce a stratified wage structure. Yet
there is considerable evidence that suggests that wage rates differ
extensively among individuals with similar endogenous productivity
characteristics. These wage differentials can be viewed as differences
in the rate of return on human capital investment.
Some of these differences may be related to traditional insti-
tutional factors such as unionization, concentration, spatial barriers
to market entry, and imperfect information. Beyond this, however,
lies the effect of social stratification on the structure of labor
supply. Rather than randomly distributed, rates of return seem to be
significantly related to race and sex. Discrimination in the labor
market can take a number of different forms each contributing in a
distinct way to differentiating rates of return.67
Wage rates (or rates of return) can differ among two individuals
who perform precisely the same job in the same firm. This type of
differential might be termed "pure wage discrimination." The more
complicated forms (and possibly the mole pervasive) involve restricted
67Thurow has attempted to catalog the several types of discrimi-
nation found in the labor market and analyze the effect of each. His
"catalog" can be found in Lester C. Thurow, Poverty and Discrimination,op. cit., p. 117.
70
53
access to particular industries, occupations, or firms. Occupational
and industrial stratification will result in different rates of return
for the same endogenous productivity characteristics and thus play a
potentially important role in the distribution of personal earnings.
How important restricted access is in determining the distribution
of rates of return can only be ascertained through empirical investi-
gation.
Social Stratification andWage Theory
Taken to its extreme,,social stratification theory stands in
direct contrast to neoclassical theory. Where the traditional analysis
focuses on maximization behavior subject to economic constraints,
stratification analysis ultimately places responsibility on the
"constraints" for determining economic outcomes. In terms of the
personal distribution of earnings, the key variables in social strati-
fication theory are beyond the control of the individual. To summarize,
they include:
(1) Opportunity for private investment in human capital(2) Opportunity for social investment in human capital(3) Restrictions to entry into specific occupations(4) Restrictions to entry into specific industries(5) Job discrimination within individual firms(6) Wage discrimination within individual jobs
By allowing these factors to enter the formulation of wage
theory, two things are accomplished. First, the barriers to mobility
stressed in the institutional analysis ("noncompeting" groups,
imperfect labor market in:ormation, unionization, and spatial
I 71
54
limitations to entry) are extended to include the obviously important
factor of discrimination in the labor market. Second, the inclusion
of the social stratification variables provides a framework for
understanding the specific distribution of jobs and earnings over the
distribution of persons. Traditional human capital theory fails to
adequately explain the distribution of endogenous productivity
characteristics while traditional institutional analyses fail to
specify which individuals will gain access to which sectors of the
economy. Social stratification theory thus may provide part of the
answer necessary to close the system used to explain personal earnings.
There is one important problem with stratification analysis,
however. Taken alone it provides no more than a description of the
wage distribution at a given point in time. In this sense it is not
a "theory" of wages per se. Its key parameters, race, sex, and social
class, are not particularly useful by themselves in analyzing changes
in the distribution of earnings. Consequently, stratification analysis
must become part of a more general theory of earnings if the theory
is to yield any more than a static description.
Toward a Complete Theory ofWage Determination
Each of the four existing theories of wages, at least in its
"pure" form, exhibits at least one critical shortcoming which prevents
it from fully explaining the observed distribution of earnings.
Marginal productivity theory fails to account for differences in
endogenous productivity characteristics and for barriers to labor
i 72
55
mobility. Human capital theory, while rectifying the problem posed
by a heterogeneous labor force, fails to pay adeq uate attention to
labor market imperfections. Institutional analysi
economic results of market imperfection, but fails t
control for differences in human capital or describe
personal characteristics responsible for differential
sub-markets. Finally, social stratification analysis,
focuses on the
o adequately
adequately the
access to labor
which neither.
assumes homogeneous labor supply or perfect labor market , is greatly
weakened by its inability to describe the dynamics of wage
determination.
Yet each theory provides a potentially vital element in the con-
struction of a general wage model. Productivity theory indicates thP
on average wages will bear a close relationship to labor's marginal
product. Further describing the supply side of labor markets, human
capital theory predicts that relative earnings will be related to
investment in endogenous productivity characteristics. Institutlo
theory poses the possibility that labor market imperfections will
impinge on the wage determination process in such a way that the
distribution of earnings (for individuals with similar human capital)
will partially reflect industry and occupational attachment. And
social stratification analysis extends the traditional institutional
analysis to account for variation in human capital investment and
different rates of return on capital due to differences in race, sex,
and social class.
nal
i 73
56
Each thus provides part of the catalogue of variables which
enter into the wage determination process. But the real problem is
determining how much of the wage distribution is best described by
each theory. The few empirical studies which have attempted to test
wage models using variables from more than one theoretical framework
have produced somewhat ambiguous results.
The most complete of these is Weiss's study of concentration
and labor earnings.68
Before controlling for personal characteristics
and other industry variables in his 1966 micro data study, Weiss
found annual earnings of male operatives in unregulated industries
to be significantly correlated with both "unionization" and concen-
tration.69 He reported that, "Unionization seems to raise annual
earnings by about 16 percent when concentration is low, but to have
no effect when CCR (concentration) is high. Concentration seems to
raise earnings by about 33 percent when unions are weak, but by only
13 percent when they are strong."70
After the addition of personal
characteristics (residence, race, age, education, family size, and
migration) and other industry variables (employment growth, size of
establishment, type of manufacturing, percent male employment, percent
68Leonard W. Weiss, "Concentration and Labor Earnings,"American Economic Review, March 1966.
69Weiss's actual regression result was:
Y = 1936 + 53.47 CCR + 23.74 U - .4426 UCCR R2= .0401
(280.5) (7.81) (4.16) (.1030) N = 5187Y = 4419
70Ibid., pp. 104-105.
i 74
57
skilled employees, percent employment nonwhite, and percent of
,'mployees residing in the South), the effects of unionization and
concentration decrease significantly. Unionization remaii,a
statistically significant (t = 5.0) but the coefficient falls to the
level where an industry that is 90 percent organized yields earnings
which are only 6-8 percent higher than an industry with only half of
the employees covered by collective bargaining agreements.
Concentration is now only barely significant (t = 2.1) and increased
annual earnings (resulting from a difference in CCR of 40 points-
20 vs. 60) amount to no more than 3 to 5 percent.71
After the
addition of the personal characteristics data, Weiss can explain about
34 percent of the variance in earnings. He concludes that, "The
effects of most industry characteristics are nonsignificant and often
of unexpected signs after personal characteristics are introduced.
In general, employers who for any reason pay high salaries receive
'superior' labor in the bargain. The general picture is one of fairly
efficiently working labor markets, even where substantial monopoly
may exist.H72
Weiss's results, however, hardly justify this optimistic con-
clusion. For one thing, Weiss specifies the "unionization" variable
at the industry level rather than at the micro level. Stafford has
shown that the average effect of union membership on relative wages is
71Ibid., p. 108.
72Ibid., p. 116.
i 75
58
10-16 percent after controlling for education, age, industry, city
size, region, and race when union membership is measured for the
individual rather than the industry.73 In addition, Johnson and
Youmans indicate that after controlling for age and education, the
effect of unicnism is actually double that found in early institu-
tional studies.74
In their study union membership increases relative
wage rates by 34.2 percent. One interpretation of this result is that
unionism is a substitute for more education. Unions insulate less
educated workers as well as younger workers from the usual effect of
education and age on wage levels. A correct specification of the
unionization variable might significantly change Weiss's results.
There are other weaknesses in the Weiss study which merit
attention. One weakness lies in his sample of industries. By
restricting his research to individuals employed in mining, con-
struction, manufacturing, transportation, communications, or public
utilities, he fails to account for the full variance in unionization
and concentration in the economy. Including workers in other services
and in wholesale and retail trade would have expanded the variance in
his industry variables and probably would have increased the signifi-
cance of these factors.75
73Frank P. Stafford, "Concentration and Labor Earnings: A
Comment," American Economic Review, March 1968.
74George E. Johnson and Kenwood C. Youmans, "Union Relative Wage
Effects by Age and Education," Industrial and Labor Relations Review,
January 1971.
75We would expect this result because it is generally known that
76
59
Even more important, the specification of the model leaves the
results ambiguous. Many of the personal characteristics in the model
have little or no relationship to endogenous productivity character-
istics and consequently muddy the interpretation that can be given to
"fairly efficiently working labor markets." The coefficient of
-681.30 on the dummy value for Negro (with t=7.3) indicates a signi-
ficant market imperfection due to some form of racial discrimination.
The same is true for the dummy value for Spanish surname.
Similarly there are a number of industry variables which are
significant and reflect market imperfections rather than differences
in the endogenous productivity characteristics in the labor force.
The sex composition of the workforce in an industry is significant
and in the extreme case of perfect sex segregation might yield a
difference in annual earnings of $619. Using a dummy variable for
durable manufacturing also makes little intuitive sense as a measure
of human capital although the coefficient is significant and accounts
for a $211 difference in annual earnings.
If these variables are considered as industry characteristics or
"stratification" factors rather than as human capital factors, the
degree of misallocation in the labor market is much greater than
Weiss 2stimates. Weiss adniits as much in his conclusion.76
the service industries and wholesale and retail trade industries arevery weakly unionized and relatively competitive at the same time
that they are generally low wage. These points should pivot the
regression line around giving a higher coefficient (higher slope) on
the industry variables.
76Weiss, op. cit., p. 115.
60
This does not necessarily imply that no misallocationresults from high-wage payments in concentrated industries.Labor "quality" in this study includes such personalcharacteristics as race, which may be quite irrelevant tothe objectively evaluated productivity of the laborer involved.It has been suggested that firms with monopoly power usepart of their profits to hire congenial or socially acceptableemployees, an option not Ivai_able to employers subject to morestringent competitive pressures. If so, the earnings of laborin monopolistic industries may still exceed its marginal-revenue product, even though they apparently approximate thevalue of its alternative product.
In this case the institutional factors explain a large part of the
variance in the personal earnings distribution after controlling for
endogenous productivity characteristics.
Two other recent studies appear to add some collaborative
evidence to this conclusion. In their study of the Chicago labor
market, Rees and Shultz controlled for age, education, experience, and
seniority.77
They found that among material handlers, the "mean wage
of nonwhites is twenty-nine cents below the mean wage of whites. . . .
The coefficient of the nonwhite dummy in the regression is a negative
thirty-one cents, indicating that only about two cents per hour
can be attributed to differences between nonwhites and whites in the
other characteristics that enter into the regression."78 The
addition of establishment variables to the Chicago labor market
regressions reduced the effect of the race dummy, but did not eliminate
its significance altogether. Such a result seems to indicate that part
77Albert Rees and George P. Shultz, Workers and Wages in an Urban
Labor Market (Chicago: University of Chicago Press, 1970).
78Ibid., p. 106.
78
61
of the variance in racial wage differentials is due to discrimination
in access to "high-wage" firms. The remaining differential arises
from wage discrimination within each firm.79
An even more recent study by Wachtel and Betsey, using multiple
classification analysis on micro data, found a large portion of the
residual variance in wage rates (after controlling for education,
experience, race, sex, age, and marital status) could be explained by
a composite "occupation-industry" variable.80
Region of employment,
city size, and union status were also significant after controlling
for personal characteristics.
While both of these studies find large wage differentials
related to labor market barriers, neither indicates precisely what
factors operate on the demand and supply sides to produce this result.
Rees and Shultz used dummy variables to account for "industry" while
Wachtel and Betsey relied on dummy variables for occupation - industry
combinations. Neither study addresses the question of what industry
factors--higher profits, concentration, restricted access, etc.- -
are responsible for the significant coefficients on "industry."
Beyond the specification problem there is an even greater
weakness in all of these recent attempts to generate wage functions:
none develop an explicit comprehensive wage theory with which to
79For similar results see Alice Kidder, "Interracial Comparisons
of Labor Market Behavior," Ph.D. dissertation, M.I.T., 1967.
80Howard M. Wachtel and Charles Betsey, "Employment at Low Wages,"
The Review of Economics and Statistics, May 1972.
79
62
interpret the reduced form results. Consequently it is difficult,
if not impossible, to disentangle the particular effects of human
capital variables, other personal characteristics, industry "ability-=to-pay" factors, industry and occupation access barriers, and "pure"
discrimination. To disentangle these factors and explore the
determinants of the personal earnings distribution requires the
development of an explicit model and a set of data which includes
specific variables for human capital factors, industry characteristics,
and stratification effects.
In essence our quest is to distinguish what forms of labor
market segmentation are responsible for the large wage differentials
we find between individual workers in the U.S. economy. Segmentation,
by our definition, is simply the division of the labor force into
"non-competing" groups for any reason: real human capital differences,
unequal access to occupations and industries, and differential wages
for precisely the same job are all bona fide forms. One particular
form of segmentation, however, is singled out for special attention.
This form is "stratification" and refers to segmentation based on
non-human capital factors. It can be said to exist whenever the labor
force is divided on the basis of race, sex, social class, or by insti-
tutional factors such as differential access to union membership.
Stratification, however, takes a number of forms itself, one of which
i.s "crowding" where workers have differential access to occupations
and industries while another is pure wage discrimination within the
same industry or occupation. Distinguishing between the effects of
1 80
63
"crowding" and "pure discrimination" is often difficult, but an
attempt at empirically isolating the two can be made.
In the following chapter, a general model of wage determination
is constructed which draws on the strengths of each theory and
indicates how the dynamics represented in each theory interact to
produce the observed wage structure. In Chapter V a part of this
theory is then tested. Attention will focus on the factors which
affect the variance in rates of return on given human capital
characteristics rather than on the process by which those character
istics are acquired. Thus the analysis will primarily attempt to
evaluate the effect of stratification on the personal earnings
distribution, given the existing distribution of human capital.
81
CHAPTER III
A GENERAL THEORY OF PERSONAL EARNINGS DISTRIBUTION
In this chapter a general theory of the personal earnings
distribution is developed. The theory is based on concepts of social
stratification, institutional economics, and human capital (which in
turn embodies the chief tenets of marginal productivity theory).
For purposes of empirical testing, a specific wage function based
on the "crowding" hypothesis is developed and a reduced form is
generated that is consistent with the overall theory. This is done
in order to test the social stratification and institutional elements
while holding constant the human capital factors.
The neoclassical formulation of economic problems normally
assumes: (1) rational individual decision-making, and (2) utility or
profit maximization subject to constraint. Both are inherent in the
marginal productivity theory and form the foundation for the human
capital approach to wage determination.) Accordingly, individuals
)In his survey article, Mincer is particularly clear on this
matter. He notes that ". . . an important attraction of this theory
is that it relies fundamentally on maximizing behavior, the basic
assumption of general economic theory." Jacob Mincer, "The
Distrib ution of Labor Incomes: A Surve:,," Journal of Economic
Literature, March 1970, p. 23. David Gordon has summarized this point
as well.
"In emphasis if not in precise substantive hypothesis, thetheories seem to suggest that individuals have a nearly
64
82
65
make two critical decisions which determine their own income. Each
worker decides how much to invest in the accumulation of human capital
stock and how much time to devote to work and how much, to leisure.
In the investment decision, individuals are constrained by innate
ability, a diminishing marginal return to increments in investment,
and an inelastic supply of investment funds. In the labor-leisure
choice, the ultimate constraint is the number of hours in the day and
the physiological need for rest. Within a broad range, individuals
determine their own wage rate, the hours of labor they supply, and
conseyently their own annual and lifetime earnings.
In the general earnings distribution theory developed here, the
neoclassical assumptions are abaadoncd. To the extent that the tra-
ditional formulation rests on the concept of free choice or "free
will," the present model end-races the opposite philosophical position;
it is at root socially deterministic. The observed distribution of
earrings is not the product of numerous personal decisions, but rather
primarily a function of social class, race, an., sex. Ultimately,
these are the exogenous determinants of wage rates. In a social
unlimited range of opportunities in the course of their
lifetimes. This implication seems to play the same role in
theories of income as the notion of "consumer sovereignty"
plays in theories of consumption and demand. In consumer
theory, that is, orthodox economists concentrate on the results
of free consumer choice among a given bundle of commodities
with different prices, rather than focusing on the ways in which
institutions tend to define or limit the bundles available for
choice."
David Gordon, Economic Theories of Poverty and Underemployment (National
Bureau of Economic Research, 1971), p. 55.
I 83
LIIIIINNOMICINIMIMINIIIIIIIIIIIIEMIIiii.
66
stratification model of earnings, individual choice or "free will"
is consigned to the error term, as it were, along with other
stochastic variables.
The distribution of earnings is a function of successive
stratification in three markets. In the human capital market, race,
sex, and social cla-is, or what may be called the "social stratifi
cation" factors, play a predominant role in the distribution of
personal investment opportunities. In the "external" labor market,
race and sex play an essential role in the distribution of individuals
across occupations and industries, human capital held constant.
Finally, in the "internal" labor market, these same social stratifi
cation factors are responsible for part of the variance in earnings
within specific occupations and industries, again assuming human
capital ceteris paribus.
The General Social Stratification Model
A simplified version of the stratification theory can be
described in a recursive system of functional equations. Race, sex,
and social class are the ultimate exogenous variables which determine
the earnings distribution through a series of primary and supplementary
transformations on human capital, occupation strata, and industry.
(I) HCi+ F
HC(R
i'Si'
Ci.,
Ai) + eHC
(II) OSik = FOS(HCi' Ai) + fos(Ri, Si, Ci) + cos
(III) INij
= FIN
(Ri
, Si) + EIN
i 84
67
(IV) Wi = FW (OSik'
INij
) + fW(HC
i'Ai) + fiW (R
i'Si) + E
w
where: HC = human capitalR = race
:; = sex
C = social classA = "innate ability"
OS = occupation stratumIN = industryW = wage ratei relates to the i
thindividual
.thij relates to the Jo, industry
k relates to the k occupation
F is a primary functionf is a se,_ondary function
f' is a tiertiary functionE is a residucal term
Equation (I) indicates the expected maximum quantity of human
capital which individual i will be able to acquire.2
Following
Becker, the quantity of human capital demanded is positively corre-
lated with ability.3 Ceteris paribus, those with greater native
ability have higher marginal rates of return (mrr) on any given level
of investment and thus have an incentive to invest more than others.
27f the present model is used to evaluate the present value
of discounted lifetime earnings, HC would refer to the expectedlifetime acquisition of human capital. If the model is used to
evaluate wage rates at a given point in time, HC ls a measure ofthe individual's human capital at the time when W is measured. To
the extent that individuals halo. different time paths of capitalacquisition, the distribution of earnings at a point in time will
diverge from the distribution of lifetime earnings.
3See Gary Becker, "Human Capital and the Personal Distributionof Income: An Analytical. Approach," W.S. Woytinsky Lecture No. 1,
Institute of Public Administration and Department of Economics,University of Michigan, 1967, p. 5.
85
68
The available supply of human capital, however, is restricted
by racial, sexual, and class discrimination in the capiLal funds
market while the demand for funds is limited by discrimination in the
labor market. In the capital market, minorities find the marginal
cost of investment funds (mcf) higher than for others. This affects
the quantity of personal human capital accumulated as well as its
structure. Entry barriers to apprenticeship programs, for example,
are equivalent to prohibitively high mcf rates for specific types of
investment.
On the demand side, external and internal labor market
discrimiacion diminish the equilibrium marginal rate of return on
investment for minorities by lowering future expected earnings.
Together, the lower mrr on investment and the higher marginal cost
of funds constrain the amount of human capital acquired by minority
members of the labor force. Human capital acquisition, according to
the stratification theory, is thus a function of innate ability
tightly constrained by the onus of race, sex, and social class origin.
As suggested earlier, the error term (eHC
) includes the effec. of
personal preference and the rational response to wage differentials
insofar as the individual is net completely constrained by other
variables in the function. This equation is less mechanistic than
may at first appear. The effect of rlce, sex, and social class
operates through cultural transference as well as through institu-
tional discrimination. Social class, for instance, obviously plays
a significant role in determining human capital investment decisions
f
VW
69
by structuring "personal preference." The same can certainly be said
for sex and probably for race. Given this perspective, "individual
decision - making" is for the rust part socially determined leaving
only a small residual which can be thought of as pure individual
personal preference. The actual "size" of this residual, of course,
is open to considerable debate.4
The second equation maintains that the probability of individual
i entering occupation stratum k is determined primarily by the
individual's stock and structure of human capital and native ability.
For present purposes, the concept of "occupation stratum" need not
be rigorously defined.5
Equation (II) can be thought of as the human capital equation
in the overall theory. At this stage, the social stratification
factors enter the equation indepeneAtly, but are of secondary
importance. Their primary role is played in the first and third
equations; that is, minority members of the labor force are assumed
to be E:reened out of certain occupations not so much because of direct
occupational entry barriers, but because of the dynamics represented
4 For an excellent discussion of how social class, family, and
school interact to determine :::e level of an individual's human capital
stock, see Samuel Bowles, "Unequal Education and the Reproduction of
the Social Division of Labor," Review of Radical Political Economics,
Fall-Winter 1971.
5For empirical purposes, an occupation stratum will lath be
defined as a set of specific census occupations which share similar
specific vocational preparation (SVP) and general educational develop-
ment (GED) requirements as reported by the U.S. Departm2nt of Labor,
The Dictionary of Occupational Titles (Washington: U.S. Government
erinting Office, 1966).
1 87
70
in Equation (I). The stochastic term (ens) accounts in part for
Personal, occupation preference after controlling for human capital
and the influence of race, sex, and class.
Unlike occupation strata, industry attachment is based on
race and sex alone (plus a stochastic factor). This formulation
follows from the fact that most industries require a broad range of
skills and combine a large number of occupations. For the purpose of
the model, the whole spectrum of occupation strata in the economy can
be thought of as being replicated in each industry, although the
relative number in each occupation stratum varies considerably. The
theory maintains that there are racial and sexual barriers which
prevent large numbers of minority members from entering certain
industries even in occupations which require relatively little human
capital or innate ability.
The error term in Equation (III) contains a n"mber of factors
beside personal preference. Limits to geographical mobility between
labor markets has some effect on constraining "industry choice,"
given regional differences in industrial structure. Cyclical factors
in the aggregate economy also affect the relative availability of
positions in different industries. In addition, pure "luck" plays
a role in industry attachment; being in she "right" personnel office
at the "right" moment may be an important factor in determining an
individual's attachment to an industry sector.
Finally in Equation (IV), the personal distribution of earnings
is described by the distribution of the labor force into occupation
88
71
and industry "slots." Knowledge of an individual's occupation stratum
and industry attachment is sufficient to define the individual's
wage within rather narrow limits. At this point, differences in
human capital and ability as well as differences in race and sex may
still have an independent effect in terms of further defining
individual earnings.
To summarize, stratification plays its primary role in determining
tne distribution of human capital. (See Figure 3.1) But it
continues to play an independent and supplementary role at every stage
Figure 3.1 A social stratification model of the
personal earnings eistributicm.
in the wage model. Race, sex, and class affect occupational
attachment independent of their effect on capital while race
and sex are also key determinants of industry attachment. Finally
89
72
these same factors, according to the theory, affect the final
distribution of wages through pure wage discriminaticn within
specific occupations and specific industries even when human capital
and ability are equal among workers.
Like all general theories, the stratification theory cannot
explain all of the variance in the earnings distribution. The error
term in Equation (IV) must account for a large number of influences
which may have only the most tenuous connection to race, sex, and
social class. To be a complete theory of wage determination, this
framework would have to be expanded in Lwo directions. First, some
attempt would be necessary to explain why stratification and discrimi-
nation play such a crucial role in the earnings distribution, if
empirically they do.6 And second, some hypothesis would be required
about the demand side of the labor market in order co explain what
appears to be a continuing'disequilibrium in terms of industry "ability
6Why labor market discrimination persists in light of itssupposed negative effect on efficiency and profits continues to be
one of the critical unanswered questIzns in modern economics. Whether
discrimination oczurs because of employer and employ ! "tastes" as
in Becker's early analysis, or discrimination is a rational statistical
response to labor market information costs as in Arrow's treatment,or whether it occ.irs because of "capitalist attempts to divide andconquer the labor force" as in some of tne radical literature cannotbe directly tested here. What can be tested is how powerful strati-fication is it terms of the earnings distribution. For backgroundmaterial on the debate, see Gary Becker, The Economics of Discrimination(Chicago: University of Chicago Press, 1957); Kenneth Arrow, "SomeModels of Racial Discrimination in the Labor Market," RAND PublicationAM-6253-RC, February 1971; and David M. Gordon, Richard C. Edwards,and Michael Reich, "Labor Market Segmentation in American Capitalism,"
mimeo, March 1973.
90
73
to pay."7 Neither of these massive efforts is undertaken here.
Rather a more specific earnings generating function is derived which
can test for the size effect of stratification on personal earnings.
The Specific Model
The general model provides a L'anework for analyzing the total
effect of capital, labor, and product market imperfections on the
distribution of wage income. However, the scope of the present study
is limited to an investigation of only one kind of imperfection.
Here we are concern2d with the extent to which barriers to occupational
and industrial access distort the wage distribution among individuals
of equal human capital endowments. For the sake of the present inquiry,
human capital acquisition is considered exogenously determined. Thus
empirical tests will be primarily restricted to Equation (IV). The
specific model is derived from the "crowding" hypothesis first
explicitly formulated by Edgeworth in 1922 and since rejuvenated by
Bergmann.8
To begin, assume a world in which there are two industrial (or
occupational) sectors and labor is homogeneous in endogenous
7One tack taken to understand the differential "ability to pay"
begins with a theory of uneven development within a dual economy.
For more on this subject, see Robert T. Averitt, The Du.l Economy (New
York: W.W. Norton, 1968) and Barry Bluestone, "Economic Crises and
the Law of Uneven Development," Politics and Society, Fall 1977.
8See F.Y. Edgeworth, "Equal Pay to Men an' Women," Economic
Journal, December 1922 and Barbara Bergmann, "The Effect on White
Incomes of Discriminat-on in Employment," Journal of Political Economy,
March/April 1971.
91
74
productivity and inelastic in supply. Furthermore assume that in
the simplest case demand conditions are identical in both sectors
so that the marginal revenue product curves are the same in Sector A
and Sector B. (See Figure 3.2) If there are no barriers to
$wage
Sector B Sector A
MRPB
MRPA
Employment in B
Figure 3.2 No "Crowding"
AEmployment in A
intersector mobility, in equilibrium an equal number of workers will
be found in both sectors (E*A= E*) and the universal market wage will
be w* = MRP*. Each worker is paid his marginal product which reflects
his endogenous productivity. If we relax the assumption of identical
MRP curves, wages will still be equal assuming a perfectly competitive
labor market.
We can now posit that for some reason firms in Sector A refuse
to employ minority workers, restricting their workforces exclusively
to white men. All other workers are forced to find employment in
Sector B. Assuming that labor force participation does not change
92
75
after segregation is imposed, the resulting wage and employment
relati4 's will be as described in Figure 3.3.
MRPB
Swage
Sector B Sector A
MRPA
Employment InE*A
Employment in A
Figure 3.3 Simple "Crowding"
The total labor force E'A ElBequals the old level EAEB, but the
imposed segregateddistribution of workers creates a wage differ-
ential of (wA-w
B). Each worker continues to be paid the marginal
product in his sector (wA = MRPA; wB = MRPB), but now there is no
correlation between endogenous productivity and relative earnings.
In Sector A, white males are paid a wage greater than their endogenous
productivity would warrant (WA > MRP* = ERP*) while in Sector B, all
minority members are paid a wage below the level that would exist in
a non-segregated economy.In this case we can say That minority
workers are "crowded" into Sector B, resulting in lower earnings.
Imperfect mobility between sectors results in a quasi-equilibrium
where wage differentials can persist and where total output is below
93
76
its full equilibrium level.9
Once crowding exists, differences in the labor demand schedules
of the two sectors can affect the earnings differential. In
Figure 3.4, Sector A is drawn so that the marginal revenue product
MRPB r" i
1e
s
s
Employment in B
$wage
wA
t::a
wB
MRP'A
MRPA
1i
EA EA-'.Employment in A
Figure 3.4 Complex "Crowding"
of labor is higher than in sector B for every equal level of employ
ment. EAEB
continues to represent the total supply of labor
(=EIEt), while minorities are limited to employment in Sector B.
Under these conditions, the wage differential will be larger. Either
because of higher marginal physical product (MPP) or higher marginal
revenue (MR) or both, workers who have access to Sector A will benefit.
9Obviously this result requires imperfection in the product market
as well. If all product markets were perfectly coapetitive, any
employer who paid a wage higher than MRP* to attract a full complement
of white male labor would shortly be forced out of the market. At a
minimur this model requires some imperfection between economic sectors.
1 94
77
If for some reason minority workers were restricted to the sector
with a higher marginal product, it is possible that demand conditions
could offset the observed effect of simple crowding. We expect that
in most cases, however, minorities will be crowded into those sectors
where demand conditions are relatively weak, thus adversely affecting
their relative wage positior. Over time there is a tendency for
simple crowding to become "complex." Industries in sector A will tend
to substitute capital for labor because wages are high, while
industries in B will substitute labor for capital. The present situ-
ation in U.S. labor markets may reflect this long-run effect.
There is no problem in generalizing this model to n sectors.
Assuming homogeneous human capital, imperfections in the product
market, and the existence of crowding, the complete earnings distri-
bution would be described by the set of quasi-equilibrium wage rates
established in each sector. Nor is there a need to speci7y perfect
segregation by race, sex, or some other non-endogenous productivity
factor for the crowding model to be perfectly serviceable. One of
the key hypotheses to be tested, in fact, is that the distribution of
earnings is a function of the degree of crowding in each occupation
and industry. Ceteris paribus, the smaller the proportion of
minorities employed in an occupation or industry. the higher the wage.
A Mathematical Treatment
The crowding hypothesis can alEo be described mathematically.
In doing so the parameters that determine wages in the presence of
market segmentation can be derived. Assume once again that human
i 95
78
capital Is homogeneous in a two sector labor market. Furthermore
assume that the marginal productivity of labor i3 a linear function
of the number of workers in each sector and is independent of the
number of workers employed in the other sector. Finally, assume that
employers are unwilling to pay a wage to each worker that exceeds
marginal productivity and that workers in a given sector refuse
employment which fails to pay them their marginal product. This
assures that sectoral wage rates will never be above nor below the
marginal revenue product in each respective sector. These last
assumptions are only included to simplify the exposition.
The model can be expressed in two linear labor demand equations:
(1) wA = aA - bAEA
(2) wB = aB - bBEB
where b = dw/dE
and one employment constraint:
(3) ET = EA + EB.
By mlking alternative assumptions about the intercept term in
sector, a., the relative slopes of the MRP curves, bi, and the
number of workers employed in each sector, Ei (determined exogenously),
measures of the wage differential between the two sectors (6 = wA - wB)
can be derived.
Four different cases of crowding can be isolated.
96
79
(I) No Crowding (with identical demand curves)
Assume aA = aB
bA = bB
ET
is mobile between sectors
In this case there are no barriers to intersectoral mobility. Any
wage differential between the two sectors will induce some workers
to move from the lower wage sector to the higher wage sector until
wage rates are equalized throughout the whole economy. In this
instance, because the demand curves are identical, employment will
be divided equally between the two sectors in equilibrium (EA = EB)
and
(4) 6 = wA - wB = (aA
- bAEA) - (aB- bBEB) = 0
(II) Simple Crowding
Assume aA = aB
bA = bB
EB > EA
Here the MRP curves are identical, but minority workers are excluded
from Sector A. Therefore,
(5) 6 (aA bAEA) (aB bBEB)
= b(EB- E
A)
_ dw(E - EA)
dE B A
97
80
In this case the total wage differential is a function of the number
of workers confined to each sector. The greater the slope of the
identical demand curves, the larger the wage differential given
EA and EB.
(III) "Complex Crowding" - Type 1
Assume aA > aB
bA = bB
EB > EA
In this case minorities are crowded into Sector B and the demand
curve in Sector A is above the schedule in Sector B. At every level
of equal employment in both sectors, the marginal revenue product in
A is greater than in B. The wage differential, 6, will then reflect
both the "supply" effect of segregation and the "demand" effect of
the vertically shifted MRP curve.
(6) 6 = (aA - aB) + b(EB - EA)
dw= (a -a ) + [--
dE(E
B- E )i
A B
In the linear model, the two effects are simply additive although
the existence of segregation is a necessary condition for the
existence of any "demand" effect.
, 98
81
(IV) "Complex Crowding" - Type 2
Assume aB > aB
bB > bA
EB > EA
In this more general case, the demand curves have different slopes
as well as different intercepts. Here MRPB has a lower intercept
and is more inelastic than MRPAat every level of equal employment
in the two sectors.
(7) 6 = (aA - aB) + (bBEB bAEA)
dw dw(aA aB) I[dEl EB (dEl
A"A]
B
This last equation can be expressed in terms of demand elasticities
by substituting (wi/Ei)/fli for the bi in each equation. Therefore,
WA148 (11 E -[(-21 n(-±-) EA(8) 6 = (a a ) EB n B EA AA B
Rearranging the terms in equation (8) yields:
(8') wA
(1 + )- w (1 + --1--)= (aA
- aB).:I
AB riB
1 99
82
(8")
wA
(aA
- aB) (1 + 1/n
B)
wB wB (1 + 1/nA
) (1 + 1/n A)
Finally, letting wB be the numeraire (wB = 1), we can derive an
expression for the relative wage, w;i.
1 + a.A
- aB+ 1/n
B(9) w'A 1 + iin
A
Thus with employment levels set exogenously, relative wages
will be a function of the loci of the respective sectoral demand
curves. More specifically, if the intercepts are equal (aA = ad,
expression (8") reduces to:
wA
(1 + 1/n )B
(10)wB
(1 + 1/nA)
and it is clear that, given intersectoral immobility, relative wages
are a function of relative employment levels and the labor demand
elasticities in each sector.
One interesting implication of the "complex crowding" model is
that in the face of intersectoral immobility, the earnings of
minorities may still be equal to or even exceed those of the dominant
employment group if the labor demand schedule in the crowded sector
is sufficiently above that in the discriminating sector. From
equation (7) it is clear that: given equal intercepts, the wage
100
83
differential is reduced to zero when the ratio of the demand slopes
is inversely proportional to the ratio of employment in the two
sectors. That is, 6=0 when be /bA = EA/EB and alt. aB. More
generally, when the intercepts are not equal, 6=0 when:
1(11) n -
B 1/nA+ aB - aA
This follows from equation (9).
While this may be only of theoretical interest, it implies that
if for some reason crowding could not be overcome, wage equalization
could still be brought about by manipulation of the derived demand
for labor in each sector of the economy. That is, if somehow the
demand curve in the crowded sector can be raised above the demand
schedule in the discriminating sector, .the wage differential can be
reduced. Increased demand in the crowded sector can thus compensate
for the earnings effect of "oversupply."
The Reduced Form
To measure the composite effect of "crowding" and differentiated
labor demand conditions, it is necessary to hold endogenous productivity
characteristics constant and investigate the remaining variance in the
earnings distribution. This is equivalent to standardizing for human
capital and then carefully measuring the composition of the remaining
wage differential, 6.
1 101
84
Assuming that endogenous productivity is measured perfectly,
a non-zero differential indicates either some degree of crowding or
an inequality between wage rates and marginal revenue products, or
in all probability, some conbination of the two. The portion of the
differential due to the relative positions of the employment suppy
curves, EA and EB, and the MRP schedules can be identified as resulting
from industry or occupational crowding. Any remaining differential
must then be due to either imperfections in labor market information
or to pure wage discrimination within industries and occupations
(assuming no measurement or specification error in the human capital
and "crowding" variables).
Obviously this course of empirical investigation is fraught
with obstacles. Controlling totally for endogenous productivity is
an impossibly difficult task. There are a myriad of individual
characteristics which enter into the composition of an individual's
endogenous productivity. Measuring even a small number of these,
independent of the price they exact in the market, requires careful
specification. Even then it is difficult to know how much of the
remaining variance may be due to unmeasured endogenous productivity
traits.10
10Is physical height, for instance, an important "endogenous
productivity characteristics" for salesmen? If it is and this par-
ticular variable is not included in the earnings generating function,
we will obviously fail to account for all the variance in salesmen's
salaries. Worse yet we may erroneously attribute some of the variance
in earnings to another variable which is covariant with height. In
this case we run the risk of.fostering a mistaken conception about
the arguments in the earnings function.
102
85
A no less difficult problem arises in identifying the labor
supply and demand schedules for each sector of the economy. Measuring
the elasticity of demand for labor in a particular industry, let alone
in all sectors of the economy, poses some severe methodological
problems. The same can be said for measuring the sectoral labor supply
curves, even accepting the simplifying assumption of zero elasticity.
Finding useful proxies for identifying the loci of the individual
supply and demand curves consequently requires some ingenuity.
Still another problem arises in specifying the functional form
of the final model. In anything so complex as wage determination,
many factors will enter interactively rather than independently.
However the relatively simple substitution of log space for even
simpler linear space may add very little power to an earnings
generating function; the actual interactions between variables may be
much more complex than log linear.11
11The size distribution of personal income in most Western
capitalist nations appears to be lognormal leptokurtic with a Pareto
upper tail. Consequently, in order to explain how this distribution
occurred, many investigators have attempted to replicate this form
through variations in a lognormal function of human capital factors.
This research has had mixed results. See, for instance, Lester C.
Thurow, Poverty and Discrimination, op. cit. In this work, Thurow
fits an equation of the following form:
cIik
= AEd.bExk
where Iik
= income for an individual with i years of education and k
years of experience; A = shift coefficient; EDi = i years of
education; Exk = k years of experience; and b and c are income
elasticities. He concludes that education interacting with years of
work experience is an important ingredient in explaining the
103
86
For all practical purposes, it is impossible to deal
definitively with these problems in an empirical analysis. It is
possible, however, to specify some of the most obviously important
human capital variables and then, using a number of carefully con-
structed industry and occupation indices, investigate the extent to
which the remaining variance in earnings can be explained by market
imperfections. Two stage equations might be used for such an analysis,
assuming that an individual's endowment of native ability and his
acquisition of human capital temporally precede his entry into a
specific occupation and industry. In this case the first equation
would specify individual earnings as some function of endogenous
productivity characteristics plus a residual term, el. The second
equation would attempt to explain El in terms of industry and occupation
variables acting as proxies for measures of labor supply and demand.
Following this procedure and assuming careful measurement of
all variables, there would be strong evidence in support of the "pure"
human capital theory if the first equation accounted for a large part
of the variance in earnings while the second equation failed to explain
much of the variation in El. Conversely, if a large portion of the
variance in earnings was explained by equation two, this would consti-
tute evidence of significant labor market imperfections. The substance
distribution of earnings. For a more general theory of comple-
mentarities among independent variables in income generating functions,
see Martin Bronfenbrenner, Income Distribution Theory (Chicago:
Aldine-Atherton, 1971), pp. 50-54.
104
87
of these imperfections could only be known if one were to have some
confidence in the proxy variables for segregation. If these variables
truly measure "oversupply" or differential demand, then significant
coefficients in the second equation are a strong indication of
"crowding" and the residual in this equation, E2, measures, at least
in part, the earnings effect of imperfections in labor market informa-
tifm and/or "pure" wage discrimination. Thus in a well-specified
system of equations it would be possible to measure (1) the effect 'If
differential endogenous productivity on the distribution of earnings;
(2) the effect of industry and occupation crowding on wage differ-
entials between individuals; and (3) the residual effect of information
imperfections and pure wage discrimination.
No matter the propriety of a two stage analysis for testing the
stratification theory, a single equation reduced form has been used
in the present research. The regression equations take the familiar
form:
Wi
= cti. + Er3ixi + E
The use of this ec ition is warranted by the relative intractability
of more complex equation systems and by the prohibitive cost involved
in actually fitting large amounts of micro data in multiple stages.
This procedure Is not unusual in that virtually all previous attempts
at measuring the determinants of earnings through large micro samples
105
88
have also relied on single regression equations.12
'13
For the same
reasons of tractability and cost, the basic equation is fundamentally
additive.14
The right side of the equation is composed of four groups or
"modules"ofX.variables. One controls for human capital; another
controls for non-monetary effects on relative wages due to working
conditions; and the last rwo are proxies for the loci of the labor
supply and demand schedules. The actual regression equations take the
linear form:
12See, for instance, Morgan, et al., Income and Welfare in the
United States, op. cit.; Weiss, "Concentration and Labor Earnings,"op. cit.; Rees and Shultz, Workers and Wages in an Urban Labor Market,
op. cit.; Stafford, "Concentration and Labor Earings: A Comment," oz.
cit.; Johnson and Youmans, "Union Relative Wage Effects by Age andEducation," op. cit.; Bennett Harrison, Education, Training, and theUrban Ghetto (Baltimore: John Hopkins Press, 1972); and Wachtel andBetsey, "Employment at Low Wages," op. cit.
13Actually, as explained later in the text, a decision rule was
followed in fitting the equations such that an approximation to asequential equations model is obtained. In effect, the human capital
variables are held constant (or nearly so) when the industry andoccupation variables are added. This is analogous to allowing theselatter variables to explain only the residual variance in the earnings
function.
141n running the regressions, several log linear forms were
tried on several sets of data. In each case, the log transformequations did not perform appreciably better than the simpler linearequations, and in a few cases they performed a bit worse. For this
reason, and also because the additive model was much easier to inter-pret and evaluate, the final regressions were run in the additiverather than interactive form. In future research I hope to experiment
with several different transformations on the raw data. These may yield
somewhat better results if a transform can be found which more approxi-
mates the actual underlying interactions between independent variables
and the true relation between independent and dependent variables.
106
89
wij.. = a +71bj HC +1:b INDijrskrs ijkrs jmrs mrs
k m
+12:b.3sSTRAT +Ebj WC + e
ijnrs prs ijprs
n p
where wijrs
= wage rate for individual i of race r and sex s
in occupation stratum j
HC = human capital characteristic k for individual iikrs
IND.. = industry characteristic m associated with theijmrs industry within which individual i is employed
STRAT.. = a measure n of industry or occupation "crowding"ijnrs for the industry or occupation within which
individual i is employed
WC.. = a measure p of working conditions in the specificijprs occupation within which individual i is employed.
= an error term
The ability to accurately estimate this set of equations depends
on the existence of a suitably large comprehensive micro data set
and an adequate specification of each module.
107
90
The Data Source and the Set ofRegression Variables15
The basic data for this study is taken from the 1967 Survey of
Economic Opportunity compiled by the Office of Economic Opportunity and
the Bureau of the Census.16
A total of some 61,000 individuals are
found in the SEO file, approximately half of which are contained in a
self-weighting sample of the United States population. The other half
of the sample is drawn from individuals living in predominantly
nonwhite census tracts. This oversample provides much better estimates
of nonwhite population parameters and consequently it is used along
with the blacks in the self-weighting sample to estimate the black male
and black female equations.17,18
15For an extended description of the data base and how it was
compiled see Appendix A.
16The Survey of Economic Opportunity is available from the SEO
Clearinghouse, Data and Compilation Center, Social Science Building,University of Wisconsin, Madison, Wisconsin 53706. More informationabout the SEO can be obtained from the Clearinghouse including codebookand user's guide.
17Comparisons of the means and standard deviations from the
nonwhite segment in the self-weighting sample and the nonwhites in thespecial oversample indicated no significant differences in terms ofall of the variables used in. this study. However, there were signi-ficant differences between the whites in the selfweighting sampleand the whites included in the oversample. For this reason, theoversample population was added only to the black equations. The
N's were already of sufficient size in the white equations and theaddition of this special sample to the black equations allowedextensive stratification of the black population without loss ofstatistical significance. The oversample is not used in the race-sexpooled regressions.
18Unfortunately, it was necessary to delete nonblack nonwhites
from the sample population. The SEO does not contain large enough
108
91
From the SE0 file, all full-time, full-year workers were
selected.19
This subsample was further refined by the elimination of
all those who either did not report a wage rate or reported that their
present job s not their "usual job." In addit corkers below
age 25 were excluded leaving a sample of predominantly prime age
individuals. The total N in the final sample is 13,896.
Data on specific occupation, specific industry, race, sex,
hourly wage rate, years of schooling completed, region at age 16,
migration from place of residence at age 16, and union membership status
were obtained for each individual from the 1967 survey. In addition
data on vocational training were available for nearly three-fourths of
the sample from the 1966 SE0 panel. Where the 1966 and 1967 SE0
individuals matched, their training data was merged onto the 1967 tape.
Industry and occupation characteristics available from a number
of macro data sources were then merged onto each individual record in
the sample. Thus each final record contained not only data on an
samples of other minorities to oermit statistical analysis. At the
same time, other minorities h- ! sufficiently different labor market
experiences that to include them with blacks would bias the empirical
results. For information on different labor market experiences ofminority groups, see Larry Sawyers, "The Labor Force Participation ofthe Urban Poor," Ph.D. dissertation, University of Mtchigan, 1969.
19Full-time, full-year represents all those who (1) reported 30or more hours of work in the week preceding the interview and(2) reported 40 or more weeks of employment in 1966. This definitionis somewhat more lenient than the normal Labor Department definition.It was used in order to take into account those who have normal full-time jobs with some degree of seasonality and those who have full-yearjobs where a full work week is somewhat less than a full forty hours,a situation which is becoming more prevalent.
109
92
individual's schooling, for instance, but also on such factors as
the profit rate and the concentration ratio in the specific three-
digit SIC industry in which the individual worked in 1967.
After merging the macro and micro data, the total sample was
stratified into occupation groups. Each of the 298 census occupations
was matched to the Dictionary of Occupational Titles yielding
unweighted average General Educational Development (GED) and average
Specific Vocational Preparation (SVP) scores for each occupation.20
From these scores, seventeen occupation gtcd.,is were formed which
were ordinally ranked according to GED and SVP. Next, in order to
create strata with sufficient sample size, groups were added together
to form five broad occupational strata. These are the strata used in
the final analysis. Each stratum contains occupations with the same
narrow GED range and (except for stratum 5) a broader range of SIP
scores. The final occupation strata include groups 1-3, 5, 6-9, 12-14,
and 15-17. Occupation group 4 was too small to be included in the
study. Occupation groups 10 and 11 include "clerical and kindred
workers, nec" and "salesmen and sales clerks, nec." Because of the
heterogeneous nature of these categories it was necessary to eliminate
them from the final analysis.21
20For detail on the Dictionary of Occupation Titles and theconstruction of the GED and SVP scores, see Appendix A.
21 Some regressions were estimated for occupation groups 10 and
11. Except for a very weak coefficient on years of schooling completed,
there were no significant results and the coefficients of determination
were always below .05.
93
In terms of general occupational descriptions, the strata include
the following types of workers.
Occupation Stratum Type of Workers
1-3 Laborers, unskilled workers, menialservice personnel
5 Operatives, semi-skilled workers, semi-skilled clerical workers, semi-skilledservice personnel
6-9 Skilled operatives, semi-skilledcraftsmen
12-14 Mechanics and technicians, skilledcraftsmen, skilled service personnel,foremen
15-17 Professionals, high-skilledtechnicians, managers, officials
This technique of occupational stratification offers a distinct
advantage over other methods of categorizing the labor force.
Ordinarily, workers are classified into one or two-digit census
occupation categories which are differentiated according to job
title rather than the presumed requirements of the job. Following
this procedure, an operative, for example, is never compared with a
given subset of clerical workers or service personnel. Yet for many
operatives, the human capital requirements assumed necessary to
perform a given job with average proficiency are similar to the require-
ments established for workers in some clerical or service positions.
By dividing the sample on the basis of GED and SVP scores rather than
job title, we are able to compare individuals who fill positions having
similar educational and vocational requirements but who are employed
94
in different cenfius-defined occupations. This allows the analysis
of earnings to be carried out for well-defined segments of the
workforce.
The final variable set was chosen from over 180 variables on
human capital, industry, and occupation and represent the closest
proxies which could he found for each of the modules.22
In many cases
specific variables were chosen in order to make the final results
comparable with previous research.
The dependent variable used throughout the analysis is hourly
earnings which is computed in the SEO from weekly earnings and weekly
hours worked. This variable may be biased by differential overtime
rates, but it is still superior to the usual measures of annual
earnings. In most cases, the hourly wage should refer to the
individual's normal wage because of the "usual" job restriction placed
on the sample. Only in the case of abnormal overtime would a problem.
arise.
The independent variables are divided into four modules.
While the modules clearly overlap in some cases, each is an attempt
to measure an identifiable force in the earnings generating function.
HUMAN CAPITAL MODULE The seven factors in the human capital module
are used to measure the effect of acquired endogenous productivity on
individual earnings. In addition, by including these variables in the
22See Appendix A for a discussion of how the data were developed
and a detailed description of each variable.
112
95
regression equation, we can hold them constant and investigate the
effect of other variables consistent with the stratification
hypothesis.
The seven human capital variables include the following:
Schooling - Formal education is measured by the commonly used
variable, years of school completed. This is a continuous variable
with the normal expectation of a positive correlation between it and
the dependent variable. In the linear additive mode, (3 can be
interpreted as the mean marginal hourly earnings expected from an
additional year of schooling.
School-South - To control for the effect of school quality on
earnings, an interaction term is used. The school-south variable
equals the years of schooling completed multiplied by a dummy variable
(=1) if living in the south at age 16. It is expected to be negative.
A
The sum of the as for the schooling and school-south variables yields
the additional earnings from a year of schooling controlled for
region. Clearly this is not an optimal quality control measure for
a number of reasons, but better measures were not available.23
23The inherent problem with the school-south variable is that
it may measure the effect of "region" per se rather than the effect
of school quality. This is particularly true if there is little
interregional migration after age 16. In this case, if the effect of
"region" operates through factors unrelated to human capital, thefinal equation will overestimate the impact of endogenous productivity
on earnings. This will, of course, favor the human capital explanation
of earnings rather than the stratification hypothesis. Ceteris paribus,
the bias in this variable is in favor of the null hypothesis that the
industry and stratification variables have no effect on earnings.
i 113
96
Geographical information on state or SMSA, which would have been
useful for merging educational resource data onto individual records,
was deleted by the Census Bureau from the SEC user tape for reputed
reasons of confidentiality.
Training - Current or previous enrollment in an institutional
manpower training program is measured by a dummy variable.24
This
variable is intended to measure specific vocational training beyond
regular schooling. According to human capital theory, its coefficient
should be posilave representing a financial return to general training.
The coefficient would be zero only if the training was financially
provided by the current employer at no expense to the worker.25
Migration - Geographical mobility is also measured by a dummy
variable which takes on the value of 1 if an individual has not changed
residence by more than 50 miles since age 16. Migration is considered
an investment in human capital insofar as it raises the marginal
product of the migrant.26 Migration, in this sense, is analogous to
24Vocational training covered by this variable includes:(1) business college or technical training (2) apprenticeship training(3) full-time company training (4) vocational training in the armedforces (5) other formal vocational training and (6) non-regular general
schooling.
25For a discussion of the theory behind specific and generaltraining, see Jacob Mincer, "On the Job Training: Costs, Returns, andSome Implications," Journal of Political Economy, Part 2, Supplement:October 1962; and Gary Becker, Human Capital, op. cit. esp. Chapter 2.
26For an excellent discussion of the human capital theory ofmigration, see Samuel Bowles, "Migration as Investment: Empirical Tests
of the Human Investment Approach to Geographical Mobility," Review of
Economics and Statistics, November 1970.
114
97
investment in schooling or training. Given rational mobility, we
expect a negative coefficient on this variable. To the extent that
migration is undertaken for non-monetary reasons or is involuntary,
it is possible that the coefficient will not be significantly
different from zero for some groups.27
Experience - No direct measure of labor force experience is
available in the SEO. As a substitute, a variable which measures
experience as a simple function of age and formal education was
created.28
This assumes that once individuals leave school, they
immediately join the workforce and work continuously thereafter. Because
of the large variance in the pattern of female labor force participation,
this is not a particularly good measure of work experience for women,
yet it may be an adequate proxy for men.
Experience, according to human capital theory, is an important
factor for it is a form of directly usable specific on-the-job training.
The experienced salesman, for example, is more productive because he
not only knows his product, but learns through experience the personal
quirks of his customers. To account for this effect, a number of
27IfIt working married women move in response to the employment
opportunities of their husbands, migration may not have a salutaryeffect on their earnings. Thus the coefficient may very well be zero
for women. This may be complicated by a racial effect for historicallynorthern migration by blacks has been beneficial, no matter the reasonsfor mobility. Thus while white women may not benefit from migration,black women (and all men) might.
28The variable was created by making "experience" = age - years
of schooling completed - 5. This is similar to the construction followedby Thurow and others in creating an "experience" variable for theanalysis of earnings functions.
115
98
previous wage studies have used proxies for experience. Most have
found a strong positive correlation between the proxy and earnings.29
However, the use of "experience" as a human capital variable
is questionable. Because of the ambiguous nature of the "experience"
variable, little can be said about the meaning of a significant
positive coefficient on this factor.30
Nevertheless we have included
the variable in the final regressions and we will normally interpret
it as though its main effect is to augment endogenous productivity.
This, of course, biases upward the total explanatory power of "human
capital" in the earnings generating function. If we were to take the
alternative interpretation of the experience variable--that experience
or seniority reflects nothing more than institutionalized pay incre-
ments based on length of service and set out in collective bargaining
agreements or offered by employers to maintain morale--it would rightly
be considered as one of the industry factors.
29Thurow has used years in the labor force as a proxy for
"experience" or on-the-job training and concludes that a large portionof the difference between white and Negro incomes can be explained bydifferences in the returns to experience. See, Lester Thurow, "TheOccupational Distribution and Returns to Education and Experience forWhites and Negroes," Federal Programs for the Development of HumanResources, Joint Economic Committee (Washington, D.C. 1968), pp. 267-84.Rees and Shultz use "seniority" as a measure of work experience andfind it to be the most significant variable in explaining the wagesof workers in their Chicago labor market study. See Rees and Shultz,
op. cit. In other studies, the variable "age" is often used as anotherproxy for work experience.
30The ambiguous nature of this variable was discussed in the
section on "Empirical Studies of the Human Capital Earnings Function"in Chapter 2.
116
99
Specific Vocational Preparation (SVP) - The final human capital
variable measures the amount of on-the-job training time required to
gain average proficiency in an individual's census occupation. The
actual variable is continuous taking the values of 1.0 through 9.0
reflecting actual training time in months and years. (See Appendix A)
It is used directly as a measure of investment in on-the-job training
and supplements the mcasure of institutional training.
Like the "experience" variable, SVP also has an ambiguous
meaning. It differs from such variables as schooling, migration, and
institutional training in that it does not occur temporally previous
to employment in an industry or an occupation. An individual must
gain access to a specific job before SVP is acquired. Thus if
individuals are barred from entering occupations which require long
training periods, the training may in fact contribute to their marginal
product, but it should not be considered an unambiguous "human capital"
factor. If stratification exists, SVP can be considered an occupation
trait such as union affiliation in a union or closed shop.
Unfortunately there is no independent measure of "native ability"
in the SE0 and consequently the final equations are less than
completely specified according to theory. To the extent that native
ability is positively correlated with acquired human capital, at least
within race and sex groups, the absense of this factor has the effect
of biasing upward the coefficients on the specified variables in the
module. The purely independent effect of native ability must then be
assigned to the error term. A critical problem arises, however, if
1 117
100
innate ability is significantly correlated with industry or occu
pational attachment independent of human capital acquisition. In this
case some of the variance assigned to industry and occupational
stratification in the regression may in fact be due to differences in
ability. While there is no concrete evidence on which to decide this
point, it seems reasonable that innate ability probably has some
independent effect on earnings within an industry or occupation, but
little effect on determining initial employment attachment. The
information costs to the employer of acquiring independent measures of
the native ability of prospective employees probably precludes the use
of such a measure in initial hiring decisions. If this is true, the
effect of native ability on earnings will appear in the error term;
it will not significantly bias the coefficients on the industry and
stratification variables.
STRATIFICATION MODULE For measures of "crowding" or segregation, we
rely on factors which affect the relative labor supply locus for each
industry and occupation. In the stratification theory, these variables
are related to race, sex, and social class. In the traditional
institutional theory, relative supply schedules are determined through
trade unionism and sometimes by other means (e.g. civil service channels).
For present purposes, measures of relative crowding by race,
sex, and union membership status are used. Labor market stratification
occurs along other dimensions as well. However measures of social
class stratification are not available and we can only speculate about
118
i101
other non-human capital characteristics used to segment the labor
force.
Although it is tempting to equate stratification with labor
market discrimination, the hypothesis under examination does not rely
on this interpretation. Stratification may occur through a sociali-
zation process and be related, at least in part, to cultural insti-
tutions and tradition. Women, for instance, may tend to stratify
themselves into certain types of "women's work." Whether this form of
stratification is "voluntary" or not depends on a whole set of subtli
psychological and anthropological questions which cannot be easily
answered.
Union Member - Trade union membership is measured as a dummy
variable for each individual in the sample. Union membership can affect
earnings in two ways; in both cases the primary effect is through the
labor supply schedule. Often in the skilled crafts, labor supply is
directly restricted through apprenticeship programs and work rules
which are maintained so as to limit the number of workers in a specific
occupation. This also appears to occur in a number of professions.
Industrial unionism, on the other hand, has the effect of restricting
employment in a given industry through its influence in setting the
quasi-reservation price of labor.31 In either case monopoly rents are
31 In the case of craft unionism, one can think of the union as
affecting the locus of a perfectly inelastic supply curve of labor,moving it leftward on the horizontal employment axis. In the case of
industrial unionism, the union affects the locus of a perfectly
elastic supply curve, moving it upward on the vertical wage axis. In
119
102
created thus making wages positively correlated with union membership,
given industry demand schedules and of course assuming that unions
are effective in limiting industry employment.32
Percent Minority-Industry (% MININD) - Data on minority
employment was merged onto the SEO sample from the 1960 U.S. Population
Census volume on "Industry Cha-racteristics.-"33
The data refer to
105 three-digit industries. The number of white females plus black
males and black females was calculated as a percentage of total employ-
ment in each industry.
This variable is used as a measure of the relative extent of
segregation in each industry. It implicitly assumes that if there
were no "crowding" there would be an equal percentage of minority
employment in each industry. Industries with relatively few minority
employees are considered relatively "uncrowded." The lack of minority
representation is assumed to be due to some form of entry barrier
which restricts labor supply along racial and sexual lines. The
setting a "minimum" wage below which no labor will be supplied, theunion in effect is setting a reservation price.
32In the long run, of course, trade unionism may also affect
earnings through the capital-labor ratio. Higher wages in the shortrun presents an incentive to the employer to increase the capital-intensity of his production process. In doing so, the marginalproduct of labor is raised, and given labor supply restrictions, thisleads to even higher earnings.
33U.S. Department of Commerce, United States Census of
Population, 1960, "Industry Characteristics," Series PC(2) 7(Washington, U.S. Government Printing Office, 1966).
120
103
percentage of minority employment will then be inversely correlated
with earnings, according to theory.
Unfortunately the mere existence of a significant negative
coefficient of %MININD, no matter how large, is insufficient to prove
the existence of a "crowding" effect. In fact there may not be a
definitive proof of crowding at all because of the difficulty in
isolating this phenomenon from pure wage discrimination.
Theoretically we can distinguish three cases. Pure discrimination
would be the rule if minority workers were paid lower wages in all
sectors while percent minority employment ( %MININD) was invariant.
Alternatively "crowding" would be the best explanation of wage
differences if minorities were segregated into some sectors of the
economy but both minorities and the dominant group (white men) were
paid identical wages whenever both worked in the same sector. Each
of these extreme cases is, of course, clear-cut. Unfortunately the
case which is more realistic is highly ambiguous as "crowding" and
pure wage discrimination probably coexist. It is because of this
"colinearity" that the two independent effects cannot be easily
identified. The best we can do is to amass as much evidence as
possible to draw the distinction knowing full-well that it cannot be
proven. The needed evidence can be gathered by carefully specifying
the estimating equations. This matter is left to Chapter V.
Percent Minority-Occupation ( %MINOCC) - Data on minority
occupational representation was merged onto the SEO sample from the
1960 U.S. Population Census volume on "Occupational
121
104
Characteristics."34
The data refer to 298 specific census occupations.
The number of white females plus black males and black females was
calculated as a percentage of total employment in each occupation.
In addition, variables were created for each minority group separately
as well as one for all females.
Analagous to the industry measure, this variable is intended to
gauge the relative intensity of "crowding" in each occupation. Again
it implicitly assumes that if there were no occupational crowding,
each occupation would have an equal percentage of minority workers.
This variable should be inversely related to earnings, but it has the
same problem of interpretation as %MININD.
INDUSTRY MODULE The five variables in the industry module reflect
an industry's "ability to pay" higher wages. "Ability" is related
to the locus of the labor demand curves in each industry and to the
potential size of producer's surplus. In each case, the variables
chosen relate to the traditional factors used in institutional
analyses of wage differentials.
Concentration (Market Power Factor) (MPF) - The measure of con-
centration used in this study is a new one developed specifically
for merging with the SEC data. Similar to the four-firm or eight-firm
concentration ratios normally used as a proxy for measuring oligopoly
34U.S. Department of Commerce, United States Census of
Population, 1960, "Occupational Characteristics," Series PC(2) 7A(Washington, U.S. Government Printing Office, 1966).
122
105
power, the variable used here is a measure of the share of industry
revenues generated by the firms with the largest assets in the
industry.35
The major difference between this "market power factor"
and normal concentration ratios is that the former has a variant
number, of firms in the "largest asset" category. Normally there are
between three and five firms, but the range for the 105 industries
used in the analysis runs from two to eleven. This is necessitated
by the data source used LO compute the variable.
This does not appear, however, to present a critical problem,
particularly since the simple correlation between Weiss's concentra
tion ratios and the MPF's for the manufacturing sector is .89.
Whatever is lost in terms of the specification is more than compensated
by the fact that the new measure can be calculated for the whole
range of industries, not just manufacturing. In this way the full
variance in "concentration" can be taken into account in the empirical
analysis.
As in most previous studies, a positive relationship between
concentration and earnings is expected. This is particularly true
where workers are organized in strong unions. Collective bargaining
power may allow employees to appropriate a share of oligopoly profits
or gain higher wages at the expense of higher consumer prices. Where
unions are weak or nonexistent, ccncentrated industries may pay higher
35See Appendix A for greater detail on the construction of this
variable.
123
106
wagei anyway in order to forestall union organizing drives or for
purposes of employee morale. Firms in highly competitive industries
are constrained in their "ability to pay" by market forces.
Union X Concentration - An interaction term is used to improve
the specification of the relationship between unionization (which
is in the stratification module), concentration, and earnings. A
negative sign is expected on the interaction term. Unionization and
concentration each affect the wage rate positively. But for a given
level of unionization, the higher the concentration ratio, the lower
the wage rate. This follows from a theory of bargaining power and
spatial limitations to firm entry. The greater the economic and
political power of management, the easier it is for management to
withstand union wage demands. Conversely, where labor is unified and
firms are relatively weak, but spatial entry barriers provide an
appropriate "ability to pay," one expects higher wages. Where strong
unions are up against powerful corporations, the ability to extract
wage increases may be diminished. The former case is often found in
construction and trucking, the latter often in durable manufacturing.36
After-Tax Profit Rate - To measure profitability, an historical
after-tax profit rate (on total assets) was computed for each industry.
36Fot more on the theory of cor.Lentration and unionism, see
Harold M. Levinson, "Unionism, Concentration, and Wage Changes: Towarda Unified Theory," Industrial and Labor Relations Review, January 1967.Weiss was the first to use such an interaction term in regressionanalysis and found a significant negative sign in some -' his equations.See Leonard Weiss, "Concentration and Labor Earnings,' .p. cit.
124
The variable is constructed sc that it measures an average profit
rate for the period 1953-1965. An historical measure is required
for while changes in relative wages may be related to current profits,
it is only logical that wage levels are related to long-run, rather
than short-run, net income.37
Capital/Labor Ratio - The capital/labor ratio is measured by
the dollar amount of depreciable assets per production worker in 1965.
It was calculated by carefully merging data from industry tax recordd
and employment and earnings data. Theoretically, the capital/labor
ratio affects the marginal physical productivity schedule in each
industry through the production function. Assuming the existence of
barriers to labor mobility and wages equal to marginal revenue product
in each sector, the higher the capital/labor ratio, ceteris paribus,
the higher the wage.
Government Demand - Public sector influences on product demand
are measured by the percentage of an industry's output purchased
by all federal, state, and local government agencies. It was computed
from the LILInpulmOutput Matrix for 1958.38 Given the size of
goyernment expenditures and its skewed distribution by industry, it is
37There is a potential simultaneity problem raised by this vari-
able for wage costs are one of the determinants of net income (i.e.
= pQ-wL-rK). To the extent that it exists, however, simultaneitybiases the results in the opposite direction from the positive
coefficient we expect.
38Adapted from the United States Input-Output Matrix-1958.
Wassily W. Leontief, "The Structure of the U.S. Economy," Scientific
American, April 1965.
125
108
theoretically possible for government to appreciably affect the
demand schedule for each industry. In effect, the marginal revenue
schedule may be higher in industries affected by government purchases.
Theoretically, a shift in government expenditures from industry A to
industry B will then affect relative earnings if labor is relatively
immobile.
There is, in addition, another explanation for a positive
coefficient on the government demand variable. The Walsh-Healy Act
and other federal and state legislation provide that government
agencies purchase only from firms which pay the "prevailing" wage or
higher. In doing this, however, the government sector may be responsi-
ble for setting higher wages in those industries where it is a major
consumer. This too would explain a positive relationship between
government demand and earnings indicating that, ceteris paribus,
government-induced employment in the private sector offers higher wages.
WORKING CONDITIONS MODULE Two variables which measure occupational
working conditions were added to the final data set in an attempt to
control for non-monetary pffects on the wage rate. Both variables
were calculated from the Dictionary of Occupational Titles. The
working conditions scores which were added to the data set represent
unweighted averages for the census occupations and were compiled from
the specific titles in a manner similar to the calculation of GED and
SVP scores. Neither is a particularly powerful measure of working
conditions, but represent the best data available at the time of the
109
original analysis.
Physical Demands - The physical demand variable measures the
physical strength required to perform a given specific occupation.
The measure categorizes occupations from "Sedentary" (=1) to "Very
Heavy" (=5) and is represented linearly. If there is, on average,
an aversion to jobs which require heavy physical effort, a positive
relationship between this variable and earnings would be expected.
In the absence of labor market stratification, workers would have to
be compensated with higher earnings in order to perform jobs which
require extraordinary physical effort.
Negative Work Traits - The mean number of adverse working con-
ditions in a specific census occupation is the other variable in this
module. Adverse working conditions refer to extremes of heat and cold,
humidity, noise and vibration, and the existence of physical or mental
hazards on the job including fumes, odors, toxic conditions, dust,
or poor ventilation. The more adverse the conditions of work,
ceteris paribus, the higher the wage necessary to induce workers into
the occupation. The specification of this variable, however, may
preclude its usefulness for one extremely adverse working condition
may require more compensation than several minor ones. Again, the
lack of an alternative data source forced reliance on this measure.
OTHER VARIABLES AND DATA The final two variables used in the analysis
are dummy measures for race and sex. As we mentioned previously,
these variables are used in the cross race-sex equations in order to
1,7
110
distinguish between the effects of "crowding" and other forms of
racial and sexual discrimination in human capital and labor markets.
Care must be taken in interpreting these two variables because of the
difference in the underlying earnings generating functions for each
race-sex group.
The original data set compiled for this analysis of the
stratification theory included virtually hundreds of variables, many
of which were slight variations of the factors included in the final
set. The final variables were selected on the basis of their per-
formance in a large number of macro regression equations. Together
with evidence from previous micro and macro studies of wage
determination, it was possible to arrive at a final set of variables
which reflected all the prime ingredients of an earnings generating
function specified in the general stratification theory. These
-variables were then used in the micro regression equations which will
be presented in Chapter V. But first we must deal with the estimation
procedure.
128
CHAPTER IV
THE ESTIMATION PROCEDURE
Translating the general stratification theory into a particular
reduced form is problematic in itself. Moving the one step further
to fitting actual regression equations poses a number of new diffi-
culties. Before dealing with the empirical results, a brief
discussion of methodology is therefore in order. In this chapter
estimation and testing procedures are developed to circumvent possible
econometric obstacles. One of these concerns the existence of potential
multicollinearity in the exogenous variables. Another is the possi-
bility of specification error.
Potential Multicollinearity
A high degree of multicollinearity is always a potentially
serious ailment in econometric analysis.1 In the present context it
1Farrar and Glauber show clearly Why a high degree of multi-
collinearity poses a serious problem in parameter estimation. In
their words:
"The mathematics, in its brute and tactless way, tells us that
explained variance can be allocated completely arbitrarily
between linearly dependent members of a completely singular set
of variables, and almost arbitrarily between members of an
almost singular set. Alternatively, the large variances on
regression coefficients produced by multicollinear independent
variables indicate, quite properly, the low information content
of observed data, and accordingly, the low quality of resulting
111
1291
112
could in fact be fatal. Linear dependence in the set of explanatory
variables would make it impossible to statistically distinguish
between the effect of the human capital variables and the effect of
the industry and stratification factors on the earnings distribution.
In this case we would be reduced to the very unsatisfactory position
of having to resort to pure a priori reasoning in order to distinguish
between the effects of the two kinds of variables. Regressing
earnings on a nonorthogonal set of independent variables would run
the risk of a serious Type I error in which we might reject a true
hypothesis about human capital or at least seriously underestimate
its impact and seriously overestimate the impact of other factors.
For this reason it is incumbent that we test the degree of collinearity
in the data set and use an estimation procedure which minimizes the
possibility of rejecting valid human capital variables which in theory
temporally precede other factors in determining earnings.
Appendix B reproduces the means, standard deviations, and the
zero-order correlation matrices (XtX) for all of the regressions in
the analysis.2
Each matrix has been analyzed for pairwise linear
parameter estimates. It emphasizes one's inability to dis-tinguish the independent contribution to explained variance ofan explanatory variable that exhibits little or no truly
independent variation."
Donald E. Farrar and Robert R. Glauber, "Multicollinearity inRegression Analysis: The Problem Revisited," Review of Economics andStatistics, February 1967, p. 93.
2We shall use (X
tX) to refer to the zero order correlation
matrix following the notation of Farrar and Glauber. (X X) is the
130
113
dependence according to a standard rule of thumb. In addition, a
stricter test for collinearity based on a modification of Fisher's
2-transformation was used to check for significant non-zero
correlation between paired independent variables.
As an example of the test results for multicollinearity, we can
look at a portion of the (XtX) matrix for white males across all
occupation strata. Table 4.1 is representative of virtually all of
the zero-order correlation matrices used in this analysis. It is
clear that this matrix passes the weak collinearity test specified
by Farrar and Glauber.3 The simple correlations between explanatory
variables never exceed an arbitrary r.. = .8 or .9. This is usually
sufficient to rule out singularity which would be manifest in a near-
zero determinant and the consequent explosion of elements of the
inverse matrix (XtX)
-1. But this weak test would certainly not rule
out the possibility of severe arbitrariness in the coefficients of
the explanatory variables or in the size of their standard errors.
The potential impact of multicollinearity on the final
regression results therefore makes an even stronger test desirable.
Modifying Fisher's z-transformation for the confidence interval of an
estimated correlation coefficient fulfills this need. This simple
algorithm tests for substantial non-zero correlation .4 Each pairwise
cross product matrix normalized (by sample size and standard deviation)
to unit length.
3Ibid., p. 98.
4Using Fisher's 2-transformation, the confidence Interval (z)
131
School
School-S
Migration
01.6
CA)
Experience
aV
SVP
MPF
UNxMPF
Profits
Union
%Min-IND
TABLE 4.1
ZERO-ORDER CORRELATION MATRIX - -WHITE
MALES/ALL
OCCUPATION STRATA
School
School-S
Migration
Experience
SVP
MPF
UNxMPF
Profits
Union
.Min -IND
1.000
.069
-.134
-.486
.329
1.000
-.090
-.074
.032
1.000
.063
-.114
1.000
-.121
1.000
rj ... .146
I.128
-.145
.082
-.217
-.007
-.013
-.053
-.061
J-.119
.030
-.047
.041
.011
tt
.105.
-.037
-.044
.062
.051
1.092
.016
.116
-.100
.003
I-.190
-.026
1.000
.547
.307
.178
-.157
1.000
.266
.782
-.162
1.000
l.199
-.133
1.000
-.120
1.000
HUMAN CAPITAL MODULE
INDUSTRY MODULE
STRAT MODULE
115
sample correlation coefficient was tested to see if it was signifi-
cantly larger than an arbitrarily low .100 at the lower bound in the
95 percent confidence interval.5
For the (XtX) matrix presented in
around a sample correlation coefficient can be calculated as:
z zz - r p
zr
where azr
can be approximated by
1
Zr
where Zr is the z-transformation on the sample correlation
coefficient
z is the z-transformation on the population correlation
coefficient
Zr is the standard deviation of the sample distribution
For the 95 percent confidence interval around the sampler1,,
zp
= zr
(1.96) azr
To modify this formula for use as a test of significant non-zero
correlation, the z-transform for an arbitrarily low correlation, z*,
is substituted for z and Fisher's equation is solved for the Iowa
bound.
z* = z* + (1.96) ar
Using Fisher's transformation table and interpolating, the lower
bound rt. can be calculated. For a fuller discussion of Fisher's
test, see Edward J. Kane, Economic Statistics & Econometrics (New York:
Harper & Row, 1968), pp. 246-47.
5For the purpose of the present analysis, a true population
A.. < .100 was considered a strong indication of linear independence in
133
116
Table 4.1, a sample correlation coefficient, according to this
collinearity test, must exceed .146 for the true population coefficient
to exceed .100 at the 95 percent level.
Applying this procedure to the correlation matrix in Table 4.1
indicates that while there are a number of instances where a sample
coefficient exceeds .146, only two of these involve a correlation between
a human capital variable and a variable in another module. In both
of these cases, the relationship is curiously inverse suggesting that
in the determination of earnings, membership in a trade union may be
a substitute for schooling and on-the-job training rather than itself
being a function of human capital.6 Adding the variable for union
membership to an equation which already includes schooling and SVP will
then not bias human capital coefficients downward.
The same test for collinearity shows some degree of linear
dependence among the variables within the stratification and industry
modules. Union membership, concentration, and after-tax profits are
the explanatory variables. While this figure is purely arbitrary, it
was purposefully set at a low level co assure a strong test of
orthogonality. As it turns out, most of the correlation coefficients
in the (XtX) matrices used in this analysis would pass this ortho-
gonality test even if the pt. were set at an even lower level. Beside
being a strong test, the modification of Fisher's z-transformation
allows a consistent test for multicollinearity throughout the whole
analysis. An rt. was calculated for each (XtX) matrix based on a
pl. = .100 and the individual pairwise sample correlation coefficients
wetle compared with these values.
6This inverse relation is fully consistent with the findings of
Johnson and Youmans in their study of the relative effects of unioni-
zation, age, and education on earnings. See Johnson and Youmans, op. cit_.
134
117
positively correlated with each other while percent minority employ
ment is inversely related to all of these. There is also a degree of
linear dependence in the human capital module. The amount of
collinearity within these mcdules indicates that it is necessary in
some cases to choose subsets of the human capital, industry, and
stratification factors to avoid the purely arbitrary assignment of
explained variance within modules. In running the actual regressions
this was often done. This pattern of linear independence between the
human capital variables and the industry and stratification factors
and some linear dependence within each module is for the most part
repeated in all of the (XtX) matrices in the analysis. IC assures a
minimum of bias in our estimates of each module but indicates that
caution must be used in interpreting the coefficients on individual
variables.
The Estimation Procedure Mechanics
To be even more certain, however, that the small amount of
intermodule collinearity does not bias the empirical results, a two
step estimation procedure was followed in calculating the regressions.
This procedure assures the integrity of the human capital variables.
The same procedure was followed in each complete regression.
The first step in the regression analysis involved running
earnings equations which only contain the human capital variables.
In each case an attempt was made to find a human capital module which
1.35
118
maximized the explained variance.7
The second step in the estimation
procedure entailed adding stratification module variables into the
regt.essiun under the strict proviso that the addition of an explanatory
variable must not destroy the "integrity" of the best fit human capital
equation. If the addition of a given stratification variable made a
human capital factor insignificant or statistically reduced its
regression coefficient significantly, the STRAT variable was removed
from the equation to assure the integrity of the HC module. After
the inclusion of any STRAT variables, the industry and working condition
factors were added again under the same human capital provision. In
this way the assumed causal priority of the human capital variables
is not violated by the effect of possible inter-module collinearity.
In every case individual variables enter the model in a causal order
suggested by Lhe general earnings *_henry.
The initial test for module "integrity" stipulated that the
addition of a STRAT or IND variable must not be allowed to reduce a
previously significant HC variable to statistical insignificance at
the .05 level. With a few important exceptions, whenever the addition
of a STRAT or IND factor wiped out the significance of one or more
human capital variables, the newly added factor was eliminated instead.
This process was necessary in instances where there was a significant
degree of collinearity as measured by the z-transformation test.
7In actuality the "best fit" human capital equation was deemed
to be that one which minimized the standard error of the regression
"tilliate(SEEmird'
106
119
The initial t-test for coefficient integrity assures the
tatistical significance of the human capital variables, but it is
incapable of checking for an absolute change in the size of the
coer-icients after STRAT and IND variables are added. Thus a second
more rigorous test was performed on the human ,:apital coefficients
which entailed applying a standard test statistic for the difference
between two means.8
t'
Estimates of t' were computed for each human capital variable when
there was any doubt about: the size of the regression coefficient in
the complete equation. With the exception of three special instances
which will be discussed in the next chapter, t' was found to be always
well below that necessary to substantiate a significant difference in
coefficients at the 95 percent confidence level. In most cases t < 1
and rarely did it exceed 1.25.
By utilizing the collinearity tests and the two step estimation
procedure, the results from the final single regression equations
approach those that would be obtained from the use of a two-stage
technique. The strict integrity of the human capital module estimates
8From William L. Hays, Statistics for Psychologists (New York:
Holt, Rinehart, and Winston, 1963), pp. 314-19.
137
120
maintained in this way allows a robust, if not overly-conservative,
test of the "crowding" hypothesis.
Problems in Parameter Specification
Controlling for collinearity in the multi-module regression
equations assures that the coefficients on the human capital variables
are not biased by the addition of industry and stratification factors.
But errors in the specification o. Al variable and the form of theA
overall equation could result in poG. estimates of $. Of particular
concern is the specification of the dependent variable and the
absence of non-linear terms and complementarities in the exogenous
variables.
The Dependent Variable - Becker and Chiswick, as well as others
who have studied human capital models, use the natural log of earnings
as the dependent variable when investments are measured in time
equivalents (e.g. years of schooling, experience, training) rather
than dollars.9
In some empirical research a lower coefficient of
determination emerges when earnings rather than the log of earnings
is regressed on schooling and experience. Nevertheless, the dependent
variable in the present analysis is the simple linear term, hourly
earnings. This is consistent with the work of Weiss, Morgan, et al.,
Hanoch, Rees and Shultz, and Wachtel and Betsey.10
9G.S. Becker and B.R. Chiswick, "Education and the Distributionof Earnings," American Economic Review, May 1966.
10See Leonard Weiss, "Concentration and Labor Earnings," op. cit.;
118
I121
Before the final analysis was attempted, a number of preliminary
regressions were prepared on individual occupation strata using the
natural log of earnings as the dependent variable. In these experi-
ments the log specification did not perform significantly better than
a linear specification. Both specifications provided similar
coefficients of determination and standard errors of the regression
estimate. In a few cases the log equations performed a bit worse
than others. For this reason, as well as for ease in evaluating the
final results, the non-log specification was retained. While these
experiments were not performed on all occupation strata or the cross
race-sex equations there does not appear to be any evidence that the
dependent variable is less well specified than in comparable studies.
The superiority of the log specification in some research
viz-a-viz the adequacy of the normal specification in the present
study may be explained by the structure of the respective analyses
and the characteristics of the labor force sample in each study.
The present analysis is primarily carried out within individual
occupation strata rather than across the whole spectrum of occupations
in the economy. It is possible therefore that a linear relationship
exists between earnings and human capital variables within a specific
stratum while the relationship is better represented by a log linear
Morgan, et al., Income and Welfare in the United States, op. cit.;
Giora Hanoch, An Economic Analysis of Earnings and Schooling,"
op. cit.; Rees and Shultz, Workers and Wages in an Urban Labor Market,
op. cit.; and Wachtel and Betsey, "Employment at Low Wages," op. r.it.
139
122
form for the economy as a whole. A move from one stratum to the next
in this case would yield a larger than linear increase in earnings
while increased human capital in any given stratum would yield only
a linear increase in wage. The difference in specification efficiency
might also be due to the fact that the present study is restricted
to full-time, full-year prime age workers who are employed at their
"usual" jobs. The relationship between human capital and earnings
may be linear for this group while the log linear relationship found
in some studies may be a function of differential attachment to the
workforce. Differences in education, for instance, mly have a larger
impact on wages between a part-time worker and a full-time worker
than between workers who share a similar attachment to the labor force.
Non-linearities in the Exogenous Variables - A number of authors
have used non-linear human capital variables to account for the
concave earnings profile normally associated with experience, age, or
seniority.11 Normally this is accomplished by running a linear term
and its square additively; evaluation of the first derivative gives
the extreme value of the function while the second derivative assures
that the extreme value is a maximum. Figure 4.1 indicates how such
a function will often appear. If the actual profile looks like AA,
it is obvious that a linear regression estimate can do little better
11For example, Johnson and Youmans use age and age
2in their
analysis of union relative wage effects by age and education. Johnson
and Youmans, op. cit. Rees and Shultz resort to the natural logarithm
of seniority to better fit this factor in an earnings function.
Rees and Shultz, op. cit.
140
$
EARNINGS
123
YEARS OF-EXPER I ENCE
Figure 4.1 The theoretical relationship
between earnings and experience
than BB unless a quadratic form is used. Obviously BB is a very poor
representation of the true relationship between earnings and experience.
Notwithstanding, no quadratic was used in fitting the final
regression equations in Lhe present analysis. Experiments on pre-
liminary equations indicated that the relationship between experience
and earnings was generally linear for the population under study.
This is not inconsistent with the non-linear profiles of previous
research for the present study sample is composed only of those who
are in prime age and working full-time full-year. This excludes those
under age 25 and for all intents and purposes those who are semi-
retired at age 65. Thus we are attempting to fit only the part of
the curve labeled A'A'. The regression line B'B' performs this task
admirably. The addition of a square or logarithmic term would in this
I 141
124
instance fail to explain any more of the variance in earnings.
It is possible that a non-linear term on SVP would have yielded
marginally better results given the scaling of this variable.
However, if there are significant diminishing returns to longer on-
the-job training, as there is for schooling, a non-linear specification
may not be superior to the one used in the analysis.12
In any case,
the amount of additional variance that might be explained by using
non-linear forms in the human capital module probably does not
seriously affect the final results given the sample population. If
anything, non-specified non-linearities in the industry and strati-
fication modules may bias the results in favor of the relative
strength of the human capital variables viz-a-viz the "crowding"
hypothesis.
Complementarities in the Exogenous Variables - A far more serious
specification error is conceivably introduced by the absence of inter-
active relations in the independent variables. The specification
used in this analysis implicitly assumes that the effect of each of
the explanatory factors is independent of all the others and that
their separate effects are strictly additive.13
'or instance, the
...The research of Giora Hanoch is responsible for identifying
tho diminishing returns to schooling for whites and non-whites in the
North and South. See Giora Hanoch, "An Economic Analysis of Earningsand Schooling," op. cit.
13The one exception to this generalization is the use of an
interaction term to specify the relationship between unionization andconcentration.
142
125
amount of experience is assumed to have no influence on the returns
to schooling and the returns to increasing both schooling and
experience are assumed to be equal to the sum of the separate returns
to increasing each variable independently.
Among others, Thurow believes that complementarities are con-
siderably important in earnings functions.14
He has argued that
Returns are not additive but multiplicative. This may beclearly seen in on-the-job experience and education. The
returns from experience depend partially on the trainee'slevel of formal education. Low education levels make sometypes of training impossible and other types expensive, butas the levels rise, training costs fall and the variety oftraining which can be given expands. These complementaritiesalso work in the opposite direction. Most jobs require someknowledge which is peculiar to the job and is not or cannotbe acquired in school. Education and experience combinedyield larger benefits than the sum of the two.
Ignoring complementarities can consequently lead to biased
estimates for factors which enter wage determination in combination
with others. This is particularly true for equations which cover the
whole occupation spectrum. Within a given occupation stratum, however,
we are in effect holding training levels roughly constant while
observing the returns to education. In this case as Thurow has noted,
the regression estimates of returns to schooling and experience are
valid within each training level.15
Insofar as the primary focus in
14Lester Thurow, Poverty and Discrimination, op. cit., p. 71.
Also Lester Thurow, "The Occupational Distribution of the Returns toEducation and Experience for Whites and Negroes," in Federal Programsfor the Development of Human Resources, A Compendium of Paperssubmitted to the Subcommittee on Economic Progress of the U.S. JointEconomic Committee, 90th. Congress, 2nd. Session (1968) Vol. 1,
pp. 267-84.15
Ibid.
143 i
PAGE 126 OF THIS DOCUMENT WAS MISSINGPRIOR TO ITS BEING SUBMITTED TO THE ERICDOCUMENT REPRODUCTION SERVICE BECAUSE ITWAS UNOBTAINABLE FROM THE SOURCE.
144
127
experience and SVP (on-the-job training) cannot be interpreted
unambiguously as human capital variables (See Chapter II). To tie
them more directly to schooling and institutional training would
seriously jeopardize the interpretation of the whole human capital
module. Furthermore, an extensive amount of expensive experimentation
would be necessary before generating the "correct" form of the inter-
action terms, particularly given the large number of exogenous
variables used in this analysis. For these reasons we have relied for
the most part on a simple linear model to test our theory. Further
research may improve our estimates, but the marginal gain does not
seem to warrant the more than marginal cost of obtaining it.
In conclusion we must use a healthy dose of pure common sense
in evaluating the final regressions. Nevertheless we can be confident
that the problems of multicollinearity and specification error do not
seriously impugn the validity of our findings, especially within
specific occupation groups. As it turns out the actual regression
results tend to be eminently reasonable as will be shown in the next
two chapters.
145
CHAPTER V
THE REGRESSION RESULTS
Having outlined a coherent theory and generated an appropriate
reduced form and a suitable estimation procedure, we are finally in
a position to investigate the empirical results. In this chapter
each of the final regression equations will be separately analyzed.
In the following chapter the regressions will be compared and
evaluated so as to identify what portions of existing wage differ
entials are due to differences in human capital versus differences
resulting from occupational and industrial stratification.
Recalling Chapter III, the reduced form to be tested is of the
general form:
w.. = a +1:b. HC +El), IND..lj TS ijjkrs ijkrs jmrs ijmrs
+Eb.STRAT.. +1:b. WC.. + ejnrs ijnrs jprs ijprs
where i, k, m, and n refer to the ith individual with the k
thhuman
capital trait in the mth industry with n
th degree of crowding and j
refers to occupation stratum, r to race, and s to sex. Individual
equations have been run for each race and sex group for each of the
128
1.4 ft
129
five broad occupation strata. In addition pooled regressions have
been estimated for each occupation stratum across race and sex groups
and for each race-sex group across all occupation strata. rinally a
"grand pooled" regression was couputed for the whole workforce
similar to regressions often found in the literature. Altogether
there are sixty final regressions excluding those where the sample
size is too small to permit statistically significant results.
(I) Stratification by occupation group, race, and sex -
Regressions stratified by j, r, and s are used to generate a series
of distinctive earnings functions for each race-sex group. Each
separate regression is generated on individuals whose particular
occupations share similar educational (GED) and vocational preparation
(SVP) requirements. These equations are especially valuable in
exploring the degree to which wage rates vary within jobs which are
narrowly defined by human capital requirements but potentially differ
in terms of industry characteristics. Accordingly the results can be
usol to evaluate the impact on personal earnings of differences in
industrial and occupational attachment within specific labor market
strata. In addition, by stratifying by occupation group it is
possible to ascertain whether specific variables in the model affect
wage rates differentially as one moves up the occupational hierarchy.
More importantly, in running separate equations for each race-
sex group, one can gather some evidence which can be used to isolate
the impact of crowding from the effect of pure wage discrimination.
A significant negative coefficient of %MININD or 7OMINOCC would be
147
130
prima facie evidence of effective "crowding." It would mean, for
instance, that black males who gained access to white male dominated
industries farLd better than their counterparts in crowded sectors.
The same would be true for a significant coefficient in the white
male equations, the interpretation being that white men who have the
misfortune of being "trapped" in minority-impacted industries bear the
onus of crowding as well. The absence of a significant coefficient
on the STRAT variables in the individual race-sex equations would tend
to weigh against the crowding hypothesis. But the case could not be
closed on this account alone. Evidence from pooled race-sex equations
would not necessarily corroborate this negative finding if the original
STRAT variables in the separate race-sex regressions were insignificant
only because of a lack of variance in these measures. This, of course,
occurs whenever there is perfect or near-perfect labor market
segregation (i.e. apartheid).
(II) Stratification by occupation group across race-sex groups -
Stratifying by race and sex therefore leads to downward biased
estimate:: of the effect of crowding as the degree of crowding
increases beyond some point In the extreme case all differences in
earnings would end up in the constant term or in differences in the
coefficients of other regressors and the impact of industry and
occupational crowding could not be directly tested. To remedy this
potential problem pooled race-sex equations are computed for each
occupation group.
148
131
Unfortunately this solution tends to do the job too well. If
the "crowding" variables are colinear with race and sex--as they
obviously are in the case of perfect segregation--then we would now
find a potential upward bias in the new coefficients. It is possible,
for instance, that earnings differences between race-sex groups are
simply the result of "pure" wage discrimination within each industry
and occupation. Crowding may then exist, but even in its absence
members of minority groups would be paid less.
In the case of perfect segregation it is therefore impossible
to determine whether crowding has anything to do with wage determination
at all. But where there is incomplete segregation--which is the more
usual occurrence--the net impact of crowding can be approximated by
running dummy variables for race and sex in the pooled equations.
Because of multicollinearity problems mentioned in the last chapter,
somewhat arbitrary regression coefficients result, but the final
dummied equations at least put a check on the possibility of
overestimating the independen: effect of industrial and occupational
crowding. The true coefficients on the STRAT variables can then be
expected to lie between the values given in the pooled regressions
with and without the race and sex variables.
(III) Stratification by race and sex across occupation groups -
The equations stratified by j are useful for measuring the effect of
industry and occupational attachment on differential earnings within
narrow GED and SVP ranges. But by their nature these equations will
normally underestimate the full impact of human capital on the total
149
132
distribution of wage rates across the whole occupation spectrum.
Increases in schooling, training, and migration are usually undertaken
to move from one occupation group to a "higher" one. To ascertain
the total human capital effect the regressions must be pooled across
the individual occupation strata. This is the third stage in the
analysis. Again wage equations are generated for each race-sex
group independently in order to account for and measure differences
in the structure of wage generating functions.
(IV) The "Grand Pooled" regressions - The final three equations
are pooled across j, r, and s and are constructed so as to yield
estimates of the full impact of both stratification and human capital
throughout the labor force. Race and sex dummies are added in the
last equation in an attempt to generate an estimate of the net relative
impact of crowding on overall earnings. These final equations must
be treated with all due caution because of the estimation procedure
used. The absence of interaction terms in the human capital module,
the linear form of the dependent variable, and the combining of all
race-sex groups in one equation must be taken into account when
evaluating these results. Nonetheless these last regressions are of
interest particularly when evaluated in light of other findings.
The Regression Results
The regressions presented in this section are the "best fit"
equations consistent with the estimation procedure outlined in the
preceding discussion. The R2s for each regression have been adjusted
150
133
for degrees of freedom and the figures in the parentheses are
t-statistics. The 95 percent confidence level has been used throughout
to measure statistical significance1
The descriptions of each
occupation stratum are based on the mean values for the variables in
the cross ra,a- equations. These can be found in Appendix B. We
begin with the lowest skilled stratum an proceed in steps to an
analysis of occupation stratahaving greater GED and SVP requirements.
OCCUPATION SI TUN -3
Jobs in the least skilled occupation stratum require no more
than a short demonstration period for the typical worker to achieve
average proficiency.2 The average worker in this group has less
than nine years of schooling and only 8 percent have any institutional
training. Yet labor force experience averages over thirty years.
In 1967 a disproportionately large percent of this stratum's workforce
was black (27%) while a full Clird (33%) were women.
Half of the workforce are members of trade unions and are
employed in industries which have on average 36 percent minority
employment. Within specific occupations the percentage of minority
employment is even larger--43.9 percent, approximately half of whom are
1Throughout the analysis there are only a few instances where
a coefficient is presented which does not meet or exceed the .05 level
of significance. These are denoted by an asterisk (*). In most of
these ,,e coefficient is significant at better than the .10 level and
the variable reduces the standard error o. the regression estimate.
In the remaining cases the coefficient is reported for comparative
purposes.
2This figure is based on Interpolation of the SVP scale.
151
134
white women. On average each production worker in this sample had
about $20,000 worth of depreciable assets with which to work,
somewhat less than the amount in other strata. Table 5.1 contains
all of the regression results for this group.
White Males - For white men the "best fit" human capital equation
contains only schooling and the interaction term school-south as
significant variables. Together the t.,:o explain 14 percent of the
variance in earnings which average $2.71 an hour. An additional year
of schooling is valued at 7.8 cents per hour if taken in the non-south.
A year of education in the south, however, adds only one cent to the
wage rate.
The addition of the significant non-human capital variables
increases the corrected coefficient of determination (R2) to .315
and reduces the standard error of the estimate (SEE) to less than
$.74 without significantly altering the coefficients in the human
capital module. Trade union membership adds $.32 to the wage rate
which repfesents a differential of approximately 13 percent over the
wage of non-union workers in this stratum. Industry segregation of the
labor force also affects earnings substantially. Ceteris paribus,
those who become "trapped" :In an industry with minority employment 10
y,reater than "average" earn $.68 less (2 X .1633 X -2.0851) than workers
in industries with minority employment one standard deviation below
the average.3 (Standard deviations are reported in Appendix B.) This
3For consistency throughout the analysis the net effect of
152
135
TABLE 5.1
REGRESSION EQUATIONS:OCCUPATION STRATUM 1-3 BY RACE AND SEX
White Male Black Male White Female Black Female Cross aaeo-Sex
Constant
HUMAN CAPITAL MODULE
2.1321 3.3632 2.1695 1.7986 .9708 2.0723 1.0822 1.8098 1.9190 3.2099
Schooling .0779 .0618 .0' .0601 .0853 .0491 .0669 .0469 .0702 .0455
(3.17) (2.74) (1 I. (4.39) (4.01) (2.65) (4.63) (3.47) (4.13) (3.62)
School-South -.0662 -.0559 -.0 c3 -.0562 --- --- -.0314 -.0157A -.0656 -.0453
(3.56) (3.28) (6.27) (5.20) --- (3.26) (1.73) (5.70) (4.98)
Training --- .4266 .3450 - -- --- .3926 .3341
--- (2.54) (2.56) --- - - --- (2.09) (2.28)
Migration - - -.3436 -.2686 --- --- -.2306 -.2734 -.2011 -.2248
(3.54) (3.37) --- --- (2.78) (3.64) (2.03) (2.92)
Experience
Specific Voc. Prep.
STRATIFICATION MODULEUnion Member .3204 --- .4023 --- .2068 --- .2756 --- .3769
--- (2.45) --- (4.56) --- (2.07) --- (3.57) --- (4.62)
Minority--Industry --- -2.0851 --- -.7213 --- -.8593 --- -1.4104 --- -1.1829
--- (4.85) --- (2.86) --- (2.97) --- (4.21) --- (4.03)
Minority--Occupation --- -.4550 --- --- -1.1527
- (2.09) --- (5.28)
% Black Male--Occupation - ----- - _
X White Female-Occupation ---- - ---
Black Female -- Occupation- - -
INDUSTRY MODULEConcentration --- .9213 --- .3992
--- (5.57) .--- (2.36)
Union x Conc. --- --- - --
After -tax Profit--- 5.0373
--- --- (2.36)
Capital/Labor Ratio- -
Government Demand --- 1.1146 - --
--- (2.11) --- - -
WORKINn CONDITIONSMODULFt
Physical Demands - -- --- --- --- --- -.1687
--- (1.96) --- - --_-_ --- (2.4S)
Negative Work Traits --- -.1324A - --
--- (1.90) --- ---
R2 .142 .315 .200 .494 .204 .511 .220 .402 .187 .521
SEE .8174 .7386 .7527 .6034 .4337 .3511 .4475 .3962 .8091 .6269
MEAN $2.71 $2.71 $2.17 $2.17 $1.74 $1.74 $Y 36 $1.36 $2.27 $2.27
N 138 138 253 253 65 65 137 137 277 777
153
136
is the first evidence of industry "crowding" affecting the distri-
bution of earnings; it will appear many times again. We might add
that this result is clearly in opposition to theories of discrimination
which posit that whites must be paid premium T:ages to work in
industries where they are forced to associate with a large number of
minority group members.4
After the stratification variables are included in the regression,
the industry variables fail to explain any additional variance in
earnings. This is consistent with the "simple crowding" hypothesis
where the locus of the demand curve is uniform across industries but
imperfections exist in the labor supply function. Since we are dealing
with white males in this instance, neither race nor sex is directly
continuous stratification and industry variables is measured over the
range ± one standard deviation (±10) about the means. (We will
refer to this measure as a "one sigma" evaluation.) This measure is
used rather than the traditional elasticity concept because it yields
a more intuitive sense of a variable's impact on earnings. The ±10
evaluation indicates the range in hourly wages earned by homogeneousworkers in industries which differ by ±10 standard deviation in con-centration, profitability, "crowding," etc. By using this type of
measure we are neither focusing on the extreme tails of the distri-
bution nor the infinitesimal marginal effect indicated by elasticities.The overall impact of the stratification and industry variables will
often be of larger magnitude than this, but seldom smaller. For a
normal distribution, two-thirds of all observations lie within la of
the mean; for many other distributions including the "pyramidal," the
uniform, and the bi-modal, a larger percentage of observations liebeyond 10 making our measure somewhat conservative. See Daniel Suits,
Statistics: An Introducti "n to Quantitative Economic Research (Chicago:
Ram: McNally, 1963) , pp. 48-51.
4For a statement of this position see Gary Becker, The Economics
of Discrimination (Chicago: University of Chicago Press, 1957) or
Kenneth Arrow, "Some Models of Racial Discrimination in the Labor
Market," RAND Publication RM-6253-RC, February 1971.
154
137
responsible for the barriers to interindustry mobility required for
a significant STRAT factor. Other factors not specified in the
equation, but probably including imperfections in labor market
information, must account for the industry distribution of white men
in this group. It is also conceivable that "lock-in" effects of
seniority and geographical immobility generate part of the wage
differential.
In the complete equation, the physical demands variable is
significant as well. But its coefficient is negative signifying that
the remuneration for heavier work is lower than for jobs requiring less
physical exertion. This inverse relationship is maintained even after
controlling for occupation stratum (GED and SVP), education, union
membership, and the race-sex composition of each industry. If this
relationship is a valid ind:..cation of the true association between
physical demands and earned income, then workers either prefer heavier
work even at the sacrifice of earnings or some workers become "trapped"
in very low wage laboring jobs and cannot easily escape to other
occupations in this or other strata. Unless we accept the implausible
first implication, this result calls into question the validity of
the "compensatory wage" theory--at least for low-skill work groups.
As it happens, the phys'cal demands variable is seldom significant in
the overall analysis and never in the more skilled occupation strata.
The seemingly counterintuitive conclusion implied by this regression
may be due to measurement error, as we noted In a previous chapter.
155
138
Black Males Within occupation stratum 1-3, schooling has an
identical impact on earnings for both black and white men. Although
the former average almost 1.4 years less schooling than their white
counterparts, an additional year of education for either is worth the
same, 7.8 cents per hour. Only in this stratum and in stratum 6-9
is there no significant difference between coefficients on schooling
for these two groups. In all of the other occupation strata the
partial on schooling is statistically greater for whites.
The equality in dollar returns to education within 0CC STRAT
1-3 would indicate a benign condition if it were not for the fact that
internal rate of return calculations show that extra schooling is not
particularly beneficial for either white or black men as long as they
remain in this stratum. For white men the internal rate of return
based on foregone income opportunity is only 1.5 percent while that
for black men is only a little better than 2 percent given lower
opportunity costs.5
Additional schooling is obviously nct the path to
5The internal rate of return calculations in this analysis are
made according to the usual formula:
nEt
C =(1 + r) t
t=0
where C represents the opportunity cost of an added year of schoolingin terms of foregone earnings, E
trepresents the additional earnings
in period t due to the added year of schooling and r equals theinternal rate of return. In these calculations the opportunity cost,C, was set equal to the annual income earned by an average individualin the occupation strata with mean years of education.
156
139
much higher lifetime incomes at least for those "trapped" within
this stratum.6
As is, the difference in schooling completed can
explain only 20 percent of the difference in mean earnings between
the two groups based on the black male equation evaluated at the mean
value of schooling completed by white men.
As in the case of white men, southern schooling adds little (in
this case, nothing) to the wage rate of black men. But unlike the
results for the white group, both institutional training and migration
have a ?urge impact on earnings. Training adds $.43 to hourly earnings
while the failure to emigrate reduces the wage by over $.34 an hour.
Training apparently permits the black worker to move out of the laborer
occupations (laborers, n.e.c. and farm laborers) into higher paid jobs
in this stratum such as ,arehousemen, metal filers, textile knitters
and loopers; and unskilled painters. Migration represents mobility to
the higher wage labor markets of the north.
The addition of the STRAT and IND modules increases the R2
to
almost .50. Union membership adds over $.40 to the wage rate; thus
the average union member in this stratum earns more than a fifth
The annual additional earnings from an added year of schooling is
assumed to be uniform from the time the individual leaves school until
he retires at age 65. In this ease education is considered a pureinvestment good and the marginal earnings profile is assumed flat. For
white men in this example, C=$5920; Et=$156; and t=49. For black men,
the opportunity cost is only $4720.
6This conclusion is fully consistent with other findings including
those of Bennett Harrison, Education, Training, and the Urban Ghetto
(Baltimore: The John Hopkins University Press, 1972) and Wachtel and
Betsey, op. cit.
157
140
(20.3%) more than the unaffiliated worker. In percentage as well as
in dollar terms, union membership is more helpful for the black worker
than the white. Stratification by industry also affects wages
significantly, although it is not as important a factor for black men.
A +l6 difference in minority employment ( %MININD) is responsible for a
$.24 difference in earnings.
The concentration ratio or "market power factor" also affects
earnings suggesting the existence of "complex crowding." A forty point
difference in the MPF (e.g. .20 vs. .60) is related to a $.36
difference in earnings. In addition, the government demand variable
is significant. Ceteris paribus, a ±10 difference in government
purchases means a $.16 wage differential. Apparently blacks do a little
better in industries subsidized by government contracts.
White Females - Schooling is the only significant human capital
variable in the white female equation; it yields approximately the
same wage increment as was found in the equations for white and black
men. This one factor is responsible for explaining about a fifth of
the variance in earnings.
The addition of the STRAT module increase the R2
to .511 and
reduces the SEE by almo6t twenty percent. Union membership and minority
employment by industry and occupation all affect white female earnings
after controlling for education. Union membership is valued at $.21
an hour yielding a percentage wage differential between union and
non-union workers approximately equal to that for white men. The ±Ia
evaluation of percent minority employment in the industry (MINIM)) and
s3
158
141
the percent minority in the occupation ( %MINOCC) yield $.30 and $.20
differentials respectively. Taken together these three variables
disclose a considerable degree of simple "crowding." if purely
additive, the three STRAT variables suggest a potential $.71 wage
differential around a mean wage of only $1 . As in the case of white
men, the industry module variables add nothing further to the explana-
tion of earnings for this segment of the labor force.
The nearly significant negative coefficient on "negative work
traits" can probably best be explained in terms of measurement error.
This is the only instance in which the coefficient is negative. In a
few cases the expected positive sign is found; in all others, with
this exception, the variable is insignificant.
Black Females - Once more the human capital module explains
about twenty percent of the variance in earnings. Schooling has about
the same dollar impact on the wage rate as it does for the other groups
in the occupation stratum. However, because of the extremely low mean
wage rate in this instance ($1.36), the rate of return on additional
schooling is greater than for any other race-sex group. In this case,
r > 4.25%. Training has no apparent impact on earnings although the
percentage of black females in this OCC STRATUM with training (7.3%)
is only slightly less than that for black men (9.5%). The other
significant variable in the module is migration. Remaining in the
same location after age 16 reduces the average wage by $.27 an hour,
similar to the effect seen for black men. In this stratum, migration
is an important human capital variable for blacks but not for whites,
159
142
most likely signifying the greater importance of emigration from the
south for nonwhite members of the labor force.
Both the stratification and industry modules add to the explained
variance in earnings boosting theR2
to .402 and reducing the SEE to
less than $.40. Union membership is worth $.28 an hour, somewhat
less than that for both groups of men but somewhat larger than the
impact of membership on the earnings of white women. Given the low
average wage rate for non-union black women in this stratum, unioniza-
tion increases the average wage almost 22 percent. This is approxi-
mately the same amount as for black men. The ±la evaluation of %MININD
results in a wage differential of $.30 an hour, identical to the
impact of industry segregation on white women and only slightly more
than the impact on black men. In addition, a ±la difference in after
tax profit rates is valued at $.18 an hour, an indication that
differences in industry demand also affects individual earnings.
For all four race-sex groups then, the stratification variables
are significant in this occupation stratum and have coefficients
of substantial magnitude after controlling for human capital. We take
this to be evidence of significant "crowding." Further analysis will
be postponed to Chapter VI.
Cross Race-Sex - Without resorting to Chow tests, it is evident
that there are some essential differences in the earnings generating
functions for the four individual race-sex groups in 0CC STRATUM 1-3.
While the same key variables are significant (schooling, union
membership, and %MININD), there are two important differences in the
160
143
regressions. In the first place migration plays a prominent role
in the functions for both black groups, but has no apparent impact
on the earnings of whites. The second difference relates to the
influence of the industry module; these variables are also significant
for the black equations, but not for the white. Differences in
industry structure apparently have no systematic effect on wage
differentials within each of the two white groups after controlling
for supply , ide stratification. On the other hand, the wage differ-
entials for both black groups reflect "complex crowding." Put
somewhat differently, differences in both supply and demand conditions
influence the earnings distribution for blacks while differences in
labor supply conditions alone appear to account for the explained wage
differentials of similarly qualified white workers.
This structural difference in the earnings functions appears to
be related to the relative variation in the underlying distribution of
industry characteristics. The significant coefficient on the government
demand variable in the black equation may be due to the fact that
the dispersion in this factor is much greater for blacks than any
other race-sex group. The coefficient of variation for black men
is 2.2866 while for white men only 1.6779. The same can be said for
the significant coefficient on after-tax profits for black women. Here
the coefficient of variation is .713 while it is no higher than .495
for any other group. Still again this holds for concentration in
the cross race-sex equation. The absence of significant coefficients
on the industry characteristics in the white equations thus may be due
161
144
to the fact that white workers are found in relatively homogeneous
industries while some blacks gain access to "permissive" economic
environments and others do not. Those who do enter the more concen-
trated, more profitable industries earn somewhat more than their
apparently misfortunate counterparts.
Although the cross race-sex equations mask these differences
in the structure of the earnings functions, they nevertheless con-
tribute to an understanding of wage determination by their ability to
estimate the impact of crowding even where segregation is near perfect.
On average across race-sex groups, schooling taken outside the
south contributes about $.07 to the wage rate per year of education,
although southern schooling is apparently worth less than one cent.
Vocational training adds $.39 to earnings while those who never migrate
from their place of residence at age 16 earn $.20 less per hour.
Altogether the human capita' variables can explain only 19 percent of
the variance in this stratum.
The addition of the remaining modules boosts the R2 to .52. The
$.38 wage increment due to Union affiliation is equivalent to an 18
percent differential between union and non-union workers, a figure
closely in correspondence with the early institutional results of
Levinson and similar to the more recent figures given by Lewis and
Stafford.7
At least for this occupation stratum, the early
7Levinson reported 14-18 percent in his early calculations;
Stafford 10-16 percent; Lewis 10-14 percent. See Harold Levinson,
"Unionism, Wage Trends, and Income Distribution, 1914-1947," Michigan
1sz
145
institutional results based on industry data were not badly biased
by the exclusion of human capital variables.
Both %MININD and %MINOCC are significant as well, together
contributing substantially to wage differentials. The ±1(7 evaluations
are worth $.44 and $.58 per hour respectively. If union membership
and the two minority employment variables are strictly additive,
market restriction induced crowding accounts for a measured wage
interval of $1.40 around a mean wage of $2.27. Differences in the
concentration ratio add another $.21 to the total measured wage differ-
ential. But how much of this is due to crowding and how much to
"pure" discrimination?
Equation (I) is the pooled regression with race and sex
variables added.
(I)
w = 3.1451 + .0452 Schoolirg -.0378 School-South + .3375 Training(3.45) (4.15) (2.37)
-.2061 Migration + .3664 Union Member -1.0019 %MININD(2.77) (4.62) (3.51)
-.7489 %MINOCC + .3887 Concentration - .1687 Physical Demands(3.27) (2.38) (2.41)
- .2439 Blackp - .3958 Fernald°(2.74) (3.86)
R2
= .558 Sr.: = .6042
Business Studies, Vol. X, No. 1 (Ann Arbor: Bureau of Business Research,Graduate School of Business, University of Michigan, 1951); FrankStafford, op. cit.; and H. Gregg Lewis, Unionism and Relative Wagesin the United States (Chicago: University of Chicago Press, 1963).
163
146
The coefficients on %MININD and %MINOCC both decline after
adding the race and sex dummies, but the fall is not especially
precipitous. The largest decline is from -1.1527 to -.7489 for the
coefficient on %MINOCC, but even this reduction is not particularly
significant. The t-statistic for a difference in the two coefficients
using the test of means is only 1.27, well below the level necessary
for a clear indication of statistical difference.
TABLE 5.2
REGRESSION COEFFICIENTS ON THE STRATIFICATIONVARIABLES IN THE POOLED OCC STRATUM
1-3 REGRESSIONS
UNION %MININD %MINOCC
Without R,S Dummies .3769 -1.1829 -1.1527(t value') (4.62) (4.03) (5.28)
With R,S Dummies .3664 -1.0019 - .7489
(t values) (4.62) (3.51) (3.27)
Reduction due toR,S Dummies .0105 .1810 .4038
2.8% 15.3% 35.0%
Union membership, %MININD, and %MINOCC are obviously not mere proxies
for race and sex, nor is the market power factor. Even after the
race and sex dummies are added, the total measured wage interval
for the stratification module is $1.12, eighty percent of its previous
value.
164
147
Together with the evidence from the individual race-sex
equations, the pooled regressions demonstrate that crowding, at least
within this one occupation stratum, is a conspicuous factor in
determining the distribution of earned income. In addition to "pure"
wage discrimination, industry crowding seems to perform an essential
function in determining wage rates for ach race-sex group, including
that of white men. Occupational segregation is also an important
factor particularly for white women. Finally unionization plays a
substantial role in the wage determination process, a finding con-
sistent with institutional analysis. Beyond these restrictions on
the supply side, the market power factor is significant suggesting
that demand side characteristics affect earnings, again in perfect
accord with traditional institutional theory.
OCCUPATION STRATUM 5
Occupation stratum 5 is .omposed mainly of semi-skilled manual
corkers. Almost two-thirds of the white men in this stratum are
sound in jobs under the single occupation title, "operatives and
kindred workers, n.e.c." Similarly 55 percent of the black men are
found in this occupation group with another 22.4 percent being
janitors and sextons. White women are less concentrated in the
operatives category; 38.8 percent are found here while another 14
percent are clerk typists, 12.7 percent are manufacturing checkers
and examiners, and 11.5 percent are assemblers. For black women,
45.4 percent are operatives. With the exception of typists, OCC
165
148
STRATUM 5 is the traditional semi-skilled blue-collar workforce.
The average full-time worker require. between one and three
months of specific vocational prepar-,,tion to perform his or her job
adequately. (SVP=2.9) The typical worker had a little more than nine
and a half years of formal schooling and almost 11 percent have parti-
cipated in some form of institutional training program. Average
experience in the labor force is thirty years. Eleven percent of this
occupation stratum is black while 39 percent is female.
The industries in which these individuals work appear in the
aggregate statistics to be similar to those in which workers in 0CC
STRATUM 1-3 are employed. A little more than half of the workfot..:e
in stratum 5 are union members while the average minority employment
in these industries was 35 percent. Each worker has slightly more
capital to work with: $24,000 vs. $20,000 in depreciable assets/
production worker in stratum 1-3. The historical average after-tax
profit rate is about .8 percentage points higher.
White Males - mile average wage rate for white males in this group
is $2.37, 16 cents higher than in occupation stratum 1-3. The "best
fit" human capital equation explains 16 percent of 11,0 variance in
earnings with schooling, school-south, and migration each contributing
to the regression. An additional year of school is valued at $.12
per hour except in the south where it returns two cents less.
Migration is worth $.18 an hour, migrants earning some 6.6 percent
more than those who have not moved since age 16.
166
149
The addition of the STRAT and Industry modules increase the R2
to .234. Unlike its positive effect on the other groups of workers
in this stratum, union membership does not appear to affect white
male earnings. Ceteris paribus, the two-fifths of white males in this
stratum who are not members of a trade union earn the same amount as
the 60 percent who are. Industry segregation, however, does have some
effect on relative earnings. The ±la evaluation of %IININD is valued
at 23 cents an hour. This is far less than in 0CC STRATUM 1-3, but
nevertheless still substantial. Concentration is the only significant
industry variable; after its addition to the equation no other industry
variables are significant at the .05 level. A similar ±10 evaluation
of concentration suggests a $.38 wage differential.
The positive coefficient on concentration in the face of an
insignificant union membership variable cannot be easily explained.
One possibility is that union membership is sufficiently colinear
with either concentration or %MININD that its real significance is
not registered in the regression.8
This hypothesis, however, is
belied by the fact that after the introduction of the human capital
module, the addition of union membership alone still does not yield a
coefficient which is significant at the .05 level. An alternative
explanation relies on the theory that relative wages are not correlated
with unionization because of the "spillover" effects or "sympathetic"
8The zero order correlation between union membership and concen-
tration is .229; between membership and percent minority employmentin an industry ( %MININD), -.244.
167
150
TABLE 5.3
REGRESSION EQUATIONS:OCCUPATION STRATUM 5 BY RACE AND SEX
White Male Black Male White Female Black Female Cross Race-Sex
Constant 1.8370 1.9263 2.1246 1.6893 1.3942 .7647 .0355 1.2803 1.8284 2.4466
HUNAN CAPITAL MODULE
Schooling .1221 .1033 .0733 .0478 .0678 .0484 .1362 .0982 .0816 .0600
(8.54) (7.33) (5.39) (4.47) (3.70) (2.97) (5.51) (4.31) (6.86) (5.61)
School-South -.0252 -.0254 -.0418 -.0242 -.0215 -.0149* --- --- -.0316 -.0251
(2.72) (2.85) (3.94) (2.92) (2.39) (1.87) --- (4.65) (4.18)
Training ---4 --- ---- -- --- .3058 .3118
--- (3.05) (3.53)
Migration -.1827 -.1540 -.3016 -.2120 --- --- -.1664 -.1349
(2.32) (2.05) (3.24) (2.91) --- --- --- (2.73) (2.54)
Experience --- --- --- .'-- .0177 .0121 - --
-- --- --- --- (3.21) (2.42) ---
Specific Voc. Prep.
STRATIFICATION MODULEUnion Member 0.00. .6240 .2525 .2720 .2625
(8.16) (3.36) (2.83) (4.68)
Z Minority--Industry -- -.8005 -.5723 -.5404 -1.1726
(2.91) (2.34) --- (1.96) (6.22)
Z Minority--Occupation0.00.0 -.4403 -.4564
- - (1.94) (2.57)
2 Black Male--Occupation - --
2 White Female -- Occupation - -- - - - -
2 Black Female--Occupation - -
INDUSTRY YODULEConcentration .6848 --- .8154 --- --- .5214 --- .4502
(4.74) --- (6.21) --- --- (2.65) --- (3.64)
Union x Conc.
After-tax Profit 000,40 10.2611 5.3120
(4.20) (2.54)
Capital/Labor Ratio .0104 .0019
(5.15) (3.80)
Government Demand 2.8019 -(2.26)
WORKING CONDITIONSMIOULKPhysical Demands --- -.2574 - --
(2.24) ---
Negative Work Traits
R2 .158 .234 .141 .495 .055 .292 .165 .378 .099 .333
SEE .8189 .7831 .7467 .5758 .7043 .6131 .6173 .5413 .8705 .7524
MEAN $2.87 $2.87 $2.39 $2.39 $2.01 $2.01 $1.86 $1.86 $1.48 $2.48
444 444 277 277 295 295 158 158 823 823
168
151
pressure of potential union organizing attempts. Non-union firms
may pay union scale to forestall organizing drives. In this case,
while unions may have an impact on absolute wage levels for all
workers, there is no discernible effect on relative inter-industry
rates. Concentration can still play a role in wage determination under
these circumstances. It measures the ability of an industry to meet
the prevailing standard set through collective bargaining in unionized
sectors of the economy.
Black Males - The structure of the regression equation for black
males in 0CC STRATUM 5 is similar to that of white males with the
exception of a significant and extremely powerful union membership
factor. The human capital equation explains 14 percent of the variance
in earnings with the same variables as in the white wale equation.
However, the effect of schooling on earnings is significantly lower
for black men. An additional year of education increments- the average
wage by only 7.3 cents compared with over 12 cents for white men.
This is a significant difference at better than the .02 level according
to the standard test for a difference in means.9 Using the internal
rate of return method presented previously, the return for white males
is approximately 3.5 percent while that of black men is less than 2.0.
As expected, migration pays off somewhat more handsomely for
black men than for whites in the same occupation stratum. Again this
is taken to reflect the importance of migration from the south.
9t' = 2.47.
169
152
Within 0CC STRATUM 5, membership in a trade union is critically
important in the earnings function for black men. On average, black
workers of this skill level who do not have access to a union earn
$.62 less an hour. Union members thus earn 30.5 percent more than
non-union workers, a percentage much larger than most institutional
estimates with the exception of those reported in the research of
Johnson and Youmans.10 Semi-skilled manual black workers apparently
are found in two kinds of industries: relatively high wage unionized
industries where they are paid wages not far below that of their white
male counterparts and relatively low wage non-union industries where
they comprise a disproportionate share of the workforce. Which industry
sector an individual can enter is crucial in determining his income.
The minority employment factor also helps to explain some of
the variance in earnings. The ±lci evaluation of 7.MININD is valued at
$.18 an hour. Concentration influences the wage rate as well. In
this instance, the ±10 evaluation results in a hefty $.48 earnings
differential for similarly qualified workers. The combined addition
of the two stratification variables and the market power factor
increases the fromfrom .141 to .495 and reduces the standard error
of the estimate from $.75 to $.58. Quite clearly the stratification
theory explains a large part of the variance for this segment of the
labor force.
10See Johnson and Youmans, op. cit.
170
lb3
White Females lhe "best fit" human capital equation does not
...-,
ce..plain much of the variance in earnings (R :-,.0.:6) for white women
in this stratum. An additional year of schooling add9 significantly
less to average earnings than it does for white men and the internal
rate of return on additional schooling is no more than that for black
men (2.0%). Schooling taken in the south ib worth only 4.5 cents per
hour per year. None of the other human capital factors are
significant at all.
Union membership is the only variable that appears in the
stratification module. Membership adds $.25 to the wage rate over
non-union workers in this group, an addition of 13 percent. The
absence of 7.MININD and %Mili0CC may be due to colinearity with the
union membership variable, but this seems unlikely given the relatively
small zero order correlations bemeen those variables:
Union Membership
° "MININD -.146
ZINO0C -.099
Concentration was highly significant when regressed alone on
white female earnings, but it consistently tended to undermine the
into ;city of the human capital module. Thus it was deleted according
to the estimation procedure and three other variables were used as
"quasi"-instruments: after tax profits, the capital/labor ratio, and
government demand. Each of thesevariables measures some facet of
"ability to pay" with the ::10 evaluations yielding wage differentials
of $.34, $.42, and $.18 respectively. Whether these effects are
171
154
strictly additive was not tested. After the introduction of the STRAT
and IND modules, the coefficient of determination rose to .292 and
the SEE declined by $.09. Once again the institutional hypotheses
appear to be valid after controlling for human capital.
Black Females - Unlike the white female results, both schooling
and experience are important variables in explaining the wage distri-
bution for black women in this stratum. Together these two explain
17 percent of the variance in earnings. Schooling is particularly
powerful adding 13.6 cents an hour to the wage rate per year of
education. This translates into a rate of return of 7 percent, much
higher than for any of the other race-sex groups. For some unexplained
reason, southern schooling does not detract from this return. Every
additional year of labor force experience also appears to augment
earnings, in this case by 1.8 cents per hour.
The stratification module is powerful as well. Both %MININD
and %MINOCC are significant factors as well as union membership.
Unionization adds about the same amount to earnings as it does for
white women, $.27. In addition, the ±la evaluations of %MININD and
%MINOCC are valued at $.22 and $.18 respectively.
As in The equations for white and black men, concentration is
also significant indicatilg a substantial degree of "complex crowding."
The ±lo evaluation of the market power factor indicates a $.30 wage
differential. Altogether, evaluation of the stratification and industry
variables suggests a $.97 wage interval around a mean of $1.86.
172
155
The physical demands variable is significant in this equation as
well, but once again its coefficient is of the "wrong" expected sign.
The negative sign may be explained by the possibility of black female
"entrapment" as janitresses. This is a particularly low wage job
which has relatively heavy physical demands although the educational
and training requirements are not especially lower than for operatives
or assemblers.
When all of the variables are added, the complete equation
explains more than twice the variance explained by the human capital
module alone. Thus even here where the human capital variables are
relatively powerful, stratification factors still play a significant
role in wage determination.
Cross Race-Sex - The human capital equation for the pooled
regression explains only 10 percent of the variance in earnings within
0CC STRATUM 5. In this regression additional years of schooling are
worth $.08 per year except in the south where they return only $.05.
Migration is also significant reflecting the importance of this
variable for both groups of men as suggested in Chapter III. In
addition, however, the training variable turns out to be powerful
($.31) and significant at more than the .01 level. Training was never
significant within the individual race-sex equations thus suggesting
the possibility that training has an effect on wages between races or
sexes but not within them.
There is a bit of evidence in the data that training opportunities
are greater for men than for women, at least in this occupation stratum.
173
156
Thirteen percent of white males in this group had some institutional
training and 12 percent of black men. However, only 7 percent of white
females and an insignificant number of black women were exposed to
vocational training. It is then possible that differences in training
within the white male group, for instance, are not important enough
to be manifest in a significant coefficient. However the difference
in training opportunities between white men and black women, for
example, may be great enough to generate the large positive coefficient
found on this variable. This conclusion is enhanced by the fact that
once the race and sex variables are added into the complete equation,
the coefficient on training falls from .3118 to .1898.11
All of this
may be taken as evidence that the "structure" of human capital endowments
is important in addition to its absolute "quantity."
Inclusion of the stratification and industry modules more than
triples the R2 and reduces the SEE by $.12. Union membership is worth
$.26 an hour while the two minority employment variables are valued
at $.41 and $.16 according to our standard ±10 evaluation. The
impact of industry segregation is thus nearly identical in both this
stratum and the lower skilled 1-3 group. This is not true for occu-
pational segregation. In the former stratum the standard evaluation
of %MINOCC furnished a $.58 wage differential. This should come as no
surprise, however. Occupation stratum 1-3 includes a broad range of
specific jobs while group 5 is overwhelmingly composed of industry
11An identical phenomenon will be fcund in the "grand" pooled
regression reported later in this chapter.
174
157
operatives. In this case we would expect to find the major differences
in earnings related to industry attachment rather than occupational
category. Occupational crowding plays a role in wage determination
even here, but apparently a minor one.
Concentration, after-tax profits, and the capital/labor ratio
are also significant in the pooled regression. Summed together the
three are worth a substantial $.64 an hour based on the standard
evaluation.
Adding race and sex dummies to the complete equation seriously
affects the coefficients on the STRAT and IND variables as well as the
value of the training parameter. Equation (II) reports these results.
w = 1.8742 + .0613 Schooling - .0205 School-South + .1898 Training(6.13) (3.60) (2.34)
- .1494 Migration + .4012 Union Member - .4138 MININD(3.03) (4.43) (2.13)
- .1944 MINOCC + .7027 Concentration - .4754 Union x Conc.(1.07) (4.56) (3.63)
+ 7.0780 After Tax Profit + .0016 Capital/Labor Ratio(3.20) (3.20)
+ 1.5487 Government Demand - .1620 Physical Demands(2.71) (2.57)
- .3246 Black - .6704 Female R2
= .433 SEE = .6955(4.09) (11.04)
After the inclusion of race and sex, three more variables become
significant while %MINOCC drops out of the equation. For one, the
union-concentration interaction term is now significant. Evaluating
175
158
this regression for union and non-union workers and at .20 and .60
concentration ratios (as was done by Weiss) elicits the impact of
unionization under competitive vs. oligopolistic conditions. It also
demonstrates the impact of the market power factor in a unionized
industry vs. an industry not covered by collective bargaining. Among
generally competitive industries with workers in 0CC STRATUM 5,
union affiliation increases the average wage by $.31 an hour or 13.7
percent according to evaluation of the regression equation at different
levels of concentration and unionization. In the more concentrated
industries union membership is capable of increasing earnings by $.11
an hour or 4.6 percent. Alternatively, greater concentration (from
.20 to .60) raises earnings by about 12.6 percent
Concentration (MPF)
20% 60%
in non-union
No union $2.23 $2.52 +12.6%
Union $2.54 $2.63 + 3.6%
+13.7% +4.6% +17.9%
industries and by 3.6 percent when a union is present. These results
are more consistent with those of Stafford than Weiss in that both
unionization and concentration are still significant after controlling
for the human capital variables.12 Workers who end up in concentrated
12Recall that in Weiss's study, the addition of personalcharacteristics to the regression all but destroyed the significance
of the concentration term. The statistically significant coefficients
176
159
unionized industries earn approximately 18 percent more than
nonunion labor in the competitive sector.
In addition to the no's significant interaction term, government
demand also affects wage determination in this stratum. A +la
evaluation of its coefficient elicits an additional $.14 difference
in earnings. Physical demands is the third newly significant variable
after the addition of race and sex. Its coefficient is negative,
possibly displaying once again the "entrapment" of black females in
low wage physically demanding jobs.
The coefficient on the race dummy is -$.32 while that on sex
is -$.67. After the inclusion of both variables the coefficient on
%MININD declines precipitously from -1.1728 to -.4138 and %MINOCC
becomes totally insignificant. Clearly only a portion of the wage
differential between race-sex groups in this stratum can be positively
identified as directly linked to industrial and occupational crowding.
Much of the differential may be due to either pure wage discrimi-
nation within specific industries or occupations or due to segrega-
tion between firms rather than between industries. The high degree
of colinearity between the dummies (i.e. sex) and the minority employ-
ment variables makes it impossible to definitively differentiate
these effects. (See Table 5.4.)
in our results can be explained by the improvement in the measurementof concentration (through the use of the "market power factor") andthe micro measuranent of union membership. See Leonard Weiss,op. cit., p. 108.
177
160
TABLE 5.4
PARTIAL (XtX) MATRIX FOR OCCUPATION STRATUM 5
Race Sex %MININD %MINOCC
Race
Sex
%MININD
%MINOCC
1.000 -.062
1.000
-.039
.451
1.000
-.060
.420
.302
1.000
Nevertheless the consistent appearance of the STRAT variables
in the individual race-sex equations suggests an extensive degree of
crowding and together with the race-sex variables demonstrates an
even larger degree of overall stratification. Within the full labor
force (at least within stratum 5) there are large wage differentials
tied to factors which measure racial and sexual discrimination--in
one form or another--after controlling for differences in human
capital endowments. In addition there is strong evidence that sub-
stantial imperfections exist within this stratum's labor market even
for white males. In this sense the traditional institutionalist and
social stratification arguments are strongly upheld by both the
individual race-sex equations and in the pooled regressions.
OCCUPATION STRATUM 6-9
Occupation stratum 6-9 is composed of a broad range of specific
occupations which demonstrate a definite distinction between "men's"
178
161
and "women's" work and "white" and "black" work in the American
economy. This particular stratum is also noted for being the most
heterogeneously &lined of the five occupation groups used in this
analysis. Each of its specific occupations falls within a narrow
range of required "general educational development" (GED) but
potentially spans a wide range of "specific vocational preparation"
(SVP) requirements. On-the-job training can range from just six
months to, in rare cases, almost ten years (see Appendix A). For
this reason many of the results are not comparable to those found
5n other more narrowly defined strata. In the whole spectrum of
occupations from least to most skilled, we will find the regression
results for this group to 13 the most anomalous.
Over 46 percent of white men in this stratum are found in just
four specific occupations: truck and tractor drivers, general (semi-
skilled) carpenters, welders, and policemen. The four most popular
occupations for black men include truck drivers, but the other three
are shipping and receiving cletlrs, seek clerks, and hospital attendants.
Almost 55 percent of all black men in this stratum are found in these
occupations.13
13This comparison probably understates the difference in occupa-
tion categories for white and black men. There is no distinctionbetween long-haul and intra-city trucking in the specific occupationcategories given by the census. If such data were available it wouldprobably indicate that white men dominate inter-city trucking whilemost black truck drivers are found on local routes. Earnings are
considerably different for the two kinds of work.
179
162
White women are found in a different set of occupations
altogether. Almost 60 percent are found in just three occupations:
sewers and stitchers, hospital attendants, :nd receptionists. More
than 74 percent of black women are consigned to jobs as hospital
attendants, practical nurses, and sewers and stitchers in the apparel
industry. This extreme occupational segregation is one of the main
determinants of wage differentials according to the regression analysis.
The average worker in OCC STRATUM 6-9 had no more formal
education than the typical worker in OCC STRATUM 5, nor any longer
labor force experience. However the specific on-the-job training
required for these occupations is somewhat greater, as we noted, taking
in most cases between six months and a year to complete and in a few
cases more. Almost 13 percent of the workforce reported enrollment
at some time in an institutional training program. Only 9 percent of
the workers in this stratum are black and only 27 percent ace female.
Again over half of the stratum's workforce are members of trade
unions, but the variance by race-sex group is extreme.14
Fifty-six
14The large variance in union membership by race and sex is
found not only in occupation stratum 6-9 but in all other strata aswell. White men are more organized than black men and both male groupsalways exceed the u:;21,,nization rates for both groups of women. The
mean union membership rates by stratum are reproduced belOw. (See
Appendix B.)Union Membership Rates (%)
Occupation Stratum 1-3 5 6-9 12-14 15-17
White Males 58% 60% 56% 44% 11%
Black Males 47 55 44 37 n.a.
White Females 32 41 43 10 10
Black Females 29 47 40 n.a. n.a.
Total Workforce 50% 52% 52% 38% 11%
180
163
percent of white men are union members; but only 44 percent of black
men; 43 percent of white women and 40 percent of black females.
The average occupation has a minority workforce of approximately 30
percent, but as expected the standard deviation for %MINOCC is as
large as the mean.15 What is ironic about OCC STRAT 6-9 is that the
segregation of the workforce appears to be so complete that the strati-
fication and industry variables are not particularly important within
individual race-sex regressions. In this case, only the pooled
regressions can uncover the effect of "crowding" due to racial and
sexual segregation. Table 5.5 contains the regressions for this
stratum.
White Males - Both the human capital equation and the complete
equation for white males have few significant variables. Only
schooling is important in the HC module and this one variable explains
only 6.3 percent of the variance. An additional year of education is
worth a relatively small $.09 an hour.
The important factor in this stratum is union membership.
Consistent with what is generally known about the specific occupations
in this stratum, unionization is worth more than $.76 an hour thus
forging a 28 percent wage differential between union and non-union
workers. In no other occupation group is union membership so important
for white men. In addition, there is a significant coefficient on the
15The actual coefficient of variation on %MINOCC is 1.08 while
that on %MININD is .63.
1.81
164
TABLE 5.5
REGRESSION EQUATIONS:
OCCUPJ.TION STRATUM 6-9 BY RACE AND SEX
;Mite Male Black Male White Female Black Female Cross Race-Sex
Constant 2.0699
HUMAN CAPITAL MODULE.0927
(4.30)
001000.
00000040
------
000000
Oa Os /0
.063
.9506
$2.96
279
1.9847
.0791(4.04)
------
---
---
.7647
(7.28)
-2.5825(2.31)
.238
.8601
$2.96
279.
2.2122
.07640.45)
-.0477(3.27)
00
-.5384
(4.18)
4004E010
------
011.0.100
.011.00.0
00 amp
------
Ow+ Oa
40 .10.00
.156
.8725
$2.36
186
1.6901
.0597
(2.96)
-.0275(2.02)
-.3526(2.93)
.5961(4.42)
41.1
0000.=
.6535(2.77)
40.01
.320
.7874
$2.36
186
1.9361
------
--
-.1917
(2.08)
00d000
-----
.0.000
.1110.01.0
110.001.
400.10.
40.00
OPOD
010.1.
- --
--
001
0000010
.040
.4733
$1.84
106
1.6422
---
-----01.M00
-.1014*(1.12)
do.*
.3328
(3.54)
00
410
.6042(2.57)
- -
4=00 00
.166
.4453
$1.84
106
1.2439
.0678
(2.29)
-.0333(2.50)
----
00011.00
00004=00
110.1000
-----
----
004040
-----
.187
.4416
$1.72
43
1.5608
.0506(2.23).
-.0317(3.18)
,10100
MIME.
00.01.
.4966(4.57)
- - -4104,00/,
-4.4168(2.86)
0001.00
000000
01.040
-6.0458(3.52)
---
.620
.3140
$1.72
43
1.2216
.0560(3.15)
-.0318(3.05)
00
.2021
(4.62)
0.0 00
------
------
---
.04000.
---
01.004.
0040.0.0
---
- --
.091
.9563
$2.60
423
2.6503
.0511
(3.45)
-.0196(2.18)
.000000.
=0110
-- -
-
.6746(8.21)
-1.2192(4.37)
-.7525(3.73)
-z.4924(2.67)
40,0/4/0
00.0000
.00.06
0000
.380
.7927
$2.60
423
Schooling
School-South
Training
Migration
Experience
Specific Voc. Prep.
STRATIFICATION MODULEUnion Member
X Minority-Industry
X Minority-Occupation
Black Male-Occupation
X White Female-Occupation
X Black Female -- Occupation
INDUSTRY MODULEConcentration
Union x Conc.
After-tax Profit
Capital/Labor Ratio
Government Demand
WORKING CONDITIONSMODUIFPhysical Demands
Negative Work Traits
R2
SEE
MEAN
N
182
AMlb..--.011111=10.-.011.
165
variable for percent black male employment in an occupation (%BMOCC).
The usual ±lo evaluation of this factor yields an added wage
differential of $.24 an hour. Together these two stratification
factors are thus valued at $1.00 or just slightly less than a third
of the average wage. After their addition, the industry module adds
nothing suggesting that differences in demand characteristics mean
relatively little if anything after supply constraints play their role.
While not an especially well-specified equation, the regression indicates
that (1) differences in human capital are not particularly important
in explaining wage differentials within this stratum and (2) that
labor market stratification is the main actor in determining the
distribution of earnings.
Black Males - The black male equation is similar to that of white
men with the exception of the. importance of migration and concentration.
An additional year of schooling is worth 7.6 cents an hour which
according to the means test is not significantly different from the
schooling coefficient for white men in this stratum.16
Again however,
as in 0CC STRATUM 1-3, the rates of return on additional schooling
are so low for both groups that this apparent equality is not
especially valuable for black men. Schooling in the south is worth
even less, yielding only $.03 an hour.
As in every other black male regression, migration is highly
significant and powerful. Those who do not migrate during their
16t' = .57.
183
166
lifetime earn $.54 less per hour. For a full-time full-year worker
this is equivalent to almost $1100 a year.
As noted previously, industry and occupation segregation by
race and sex can be so extensive that %MININD and %MINOCC would not
be significant variables within !ndividual equations. This is
apparently true for the black male regression. Neither of these
variables is significant. However, union membership is, and once
again as in OCC STRATUM 5 its effect is robust. There is a $.60
wage gap between union and non-union workers which represents a 28
percent differential exactly the same as for white men. This is
roughly equivalent to the difference in wages found between a black
unionized maintenance painter ($2.75) and a skilled non-union hospital
attendant ($2.15).
After the inclusion of union membership, concentration adds to
the wage differential as well suggesting a "ase of complex crowding.
Unlike for white men, industry demand characteristics apparently
affect relative wages. The ±la evaluation is worth 35 cents an hour.
With unicnization and concentration included in the regression, the R2
more than doubles to .320 and the SEE declines by $.085.
WhiLe Females - The human capital equation for white women in
this group explains only four percent of the variance and neither
education, training, experience nor specific vocational preparation is
significant in this regression. Only migration is significant and
its coefficient is a relatively small -$.19. The addition of union
2membership and concentration raises the R to .166, although this occurs
184
167
only after the estimation rule concerning the integrity of the human
capital module is violated. Union membership is worth $.33 an hour
or 19.7 percent. The ±la evaluation of the concentration ratio is
valued at 23 cents an hour. Although this regression is not
particularly well fitted, it suggests that within this stratum almost
all of the explained variance in earnings is due to industrial and
occupational attachment. This is not an unreasonable conclusion given
the heterogeneous set of "women's" occupations represented in STRATUM
6-9, all of which have very similar GED and SVP ratings but differ
in terms of industrial characteristics.
Black Females - In contrast to the white female equation, the
regression for black women is quite servicable. The human capital
module explains almost 19 percent of the variance with the two variables
schooling and snocl-south. An additional year of education is worth
6.8 cents excepting the south where its value is less by 3.3 cents.
The addition of the STRAT and IND modules increases the toto
.620, the highest of any equation in the analysis. The standard error
of the estimate is only $.31 after the introduction of these modules.
Union membership is worth almost $.50 an hour. Th's represents a 33
percent differential between union and non-union s,orkers. A further
indication of the effects of labor market crowding is found in the
±la evaluation of percent black female employment In as occupation
(%BFOCC). This yields an additional differential of $.30. For
occupations that are even more "impacted" with black females the
effect is, of course, larger. The wage rate for hospital attendants,
t85
168
for example, is $.57 an hour lower than in occupations with only the
average percentage of black females. Quite obviously occupational
"crowding" severely affects earnings in this stratum.
In the industry module there is an unexpected sign on the
government demand variable. Here we find that a tla increase in
government demand apparently lowers the wage rate of black women by
nearly $.19 an hour. This is the only instance in the entire analysis
when a statistically significant government demand variable had a
negative coefficient. This rather puzzling result was found to be
merely a function of the peculiar industry composition in the micro
data for this regression.17
No other industry variables were signifi
cant so that our overall assessment is one of "simple"--but extensive--
crowding.
17Since we do not have a measure of government demand for the
"hospital" industry in our data set, the coefficient cannot be dueto the lower wages paid to hospital attendants in government subsidizedhospitals. Another solution was necessary. The puzzle was finallysolved after perusing the original data on government purchases byindustry. The weighted mean level of government expenditures in theindustries in this regression was 2.8 percent of gross sales. Thereis one industry in the whole data set which sells a larger percentageof its product to the government and also emp.,ys a relatively largenumber of black women in this o:cupation stratum. This industry is"Miscellaneous Fabricated Textile Products" employing a large numberof sewers and stitchers. It sells over 4.5 percent of its annualproduction to government agencies. With a low market power factor(.1043), a relatively low profit rate (2.5%), and an extremely highproportion of minority workers (60.7%), the average wage rate in theindustry is relatively low ($2.10 an hour in 1967). Black women,according to the data, have little access to other industries whichhave large government contracts, but also higher average wages.Consequently, in this single case, there is a negative relationshipbetween government purchases and individual earnings. This result isnot inconsistent with the general stratification theory.
186
169
Cross Race-Sex - Because of the extensive occupational segre-
gation in this stratum by both race and sex, there is a strong tendency
for the individual equations to underestimate the impact of crowding
on the distribution of earnings. Consequently the pooled regressions
may give better estimates of its effect.
In these equations the human capital variables are responsible
for explaining only 9.1 percent of the variance in earnings. Formal
schooling has a relatively weak impact on wages, an additional year
addino only 5.6 cents to hourly earnings and less than half as much
if the schooling was completed in the south. However, for the first
time in the analysis, specific vocational preparation is significant
suggesting the critical nature of on-the-job training in the occupations
within this group. The fact that SVP is statistically significant
in the pooled regression while insignificant in each of the individual
equations suggests that access to jobs wilich provide apprenticeship
(or other forms of investment in OJT) is linked directly to race and
sex. White males have freest access to training while the other race-
sex groups are provided with a lower average level of SVP.18
18The mean SVP scores for each race-sex group are:
SVP asvp
White Males 4.79 1.15
Black Females 4.37 .82
White Females 4.35 .78
Black Males 4.33 .93
Crcss Race-Sex 4.65 1.07
IS?
170
The addition of the STRAT variables compromises the integrity of
the human capital module in that SVP now disappears from the
regression. In this case we have permitted this to occur for while
the letter of the estimation procedure is violated, we feel its
"spirit" is not. The original intent of using a surrogate "two-stage"
regression was to account for the sequential relationship between an
individual's acquisition of human capital and his or her subsequent
attachment to a specific industry and occupation. In the special case
of SVP the presumed causal ordering is often the reverse. On-the-
job training such as apprenticeship is only available after a worker
has gained access to a particular job or occupation. If access is
denied on the basis of race or sex, as it often appears to be in this
group of occupations, then SVP acts more as a proxy for stratification
than a traditional human capital factor. Not to violate the original
estimation procedure under these circumstances would lead to
seriously downwali biased estimates of the effects of industrial and
occupational segregation.
Before the race and sex factors are added to the complete
equation, the introduction of the stratification module boosts the R2
to .38 and lowers the SEE by over 16 cents. Union membership alone
is worth $.67 an hour; those in organized industries or occupations
consequently earn a full 30 percent more than those who are not. In
addition, %MININD, %MINOCC, and %BMOCC are all highly significant.
If strictly additive, the sum of the ±la evaluations is worth over
$1.30 an hour, half the mean wage rate. Although quite hefty, .his
188
171
result is by no means unreasonable. It constitutes a differential not
much larger than the actual differences in mean earnings between white
men and white and black women. After the stratification module
variables are added, none of the industry variables are significant.
"Simple" crowding appears here to be the rule.
As expected the introduction of the race and sex factors severely
reduces the strength of the variables in the stratification module
with the exception of unionization. With the extreme occupation
segregation found in this stratum, the complete pooled regression
attributes much of the variance in earnings due to what we feel is
occupational attachment to the micro measured race and sex dummies.19
Equation (III) is the "best fit" complete regression including these
variables.
(III)
w = 1.8476 + .0532 Schooling + .0799 SU) + .6963 Union MilLbership(3.74) (2.23) (9.22)
- .7200 %Mi:tIND - .5476 Black - .6763 Female(2.90) (4.18) (5.41)
2R = .421 SEE = .7659
Union membership continues to be worth almost $.70 an hour clearly
labeling access to a un_o%ized occupation as the surest admission
ticket to higher earnings in this stratum, afT controlling for human
19The zero order correlations for %MINOCC and sex = .810
%BMOCC and race = .560
189
172
capital. SVP is significant once again, but use of the difference
in means test indicates that its coefficient in the final equation is
statistically lower than in the equation which contains only the human
capital variables. (C=2.16,
All of these results are consistent with what we know about the
specific occupations within this 0CC STRATUM. The Teamsters' and the
car: enters' union, for instance, have historically won wage packages
which are among the highest within the occupation spectrum. Workers
who gain access to these jobs or the other crafts gain from the
collective bargaining efforts of their unions; those who for one
reason or another do not enter these occupations will normally earn much
lower wages, ceteris paribus. This particularly affects the earnings
of women, but also limits the earning power of black men. The effect
of stratification, either in the form of "crowding" or -ure discrimi-
nation, appears to reach its peak in this stratum. Again, however,
because of the high degree of segrJgation, it is statistically
impossible to differentiate Jetween the two forms. Differences in
human capital--with the exception of SVP--only account for a small
fraction of the explained variance in earnings.
OCCUPATION STRATUM 12-14
For men, 0CC STRATUM 12-14 primarily contains individuals who
are defined by the Census as "craftsmen, foremen, and kindred workers."
The largest specific occupation for white men if "foremen" (17.7%)
followed by "mechanics and repairmen, n.e.c." (14.5%) and "machinists"
190
173
(8.5%). "Linemen and servicemen," "plumbers and pipefitters," and
"electricians" are also members of this group. For black men, the
largest single group is "mechanics and repairmen, n.e.c." (19.5%)
followed by "auto mechanics" (14.5%) and "foremen" (12.1%).
In contrast to both groups of men, most white women in this
occupation group are found in white collar jobs; almost two-thirds
(64.7%) are classified as "secretaries." Much smaller percentages are
found as medical and dental technicians and department heads and
buyers in retail outlets. The number of black women in this stratum
is so small that no statistically significant results could be obtained.
The typical worker in this category had eleven years of schooling
(10.8 years) and over 30 percent reported receiving some form of
institutional training. White females had, on average, more schooling
(11.4 years) but only 21 percent reported previous enrollment in a
vocational education program. The average occupation in this stratum
requires between two and four years of on-the-job training according
to the SVP scale. A few jobs require more.
Consistent with the specific occupational composition of this
stratum, a much smaller percentage of workers are trade union members
(38.4%). Foremen have been discouraged from organizing into unions
since Taft-Hartley and few secretaries and mechanics are members.
Only about 4 percent of the stratum is black and only 15 percent are
women reflecting the dominance ,f white men in these higher level
occupations. Table 5.6 contains the regression results for this
stratum.
191
4
174
TABLE 5.6
REGRESSION EQUATTONS:
OCCUPATION STRATUM 12-14 BY RACE AND SEX
White Male Black Male White Female Black Female Cross Race-Sex
Constant -.2914 .2250 1.9477 1.9279 1.0390 -.0761 --- -2.9893 -.7516
RUYAN CAPITAL MODULESchooling .1593 .1244 .0719 .0545 .1152 .1068 SAMPLE SIZE 4560 .1409
(8.35) (6.51) (2.76) (2.34) (2.66) (2.52) (8.91) (8.49)
School-South -.0397 -.0385TOO SMALL
-.0407 -.0370(4.58) (4.64) --- MINEN ---
FOR.(5.15) (5.00)
Training --- .5840 .4125 ---
--- (2.65) (2.14) --- --- SIGNIFICANT
Migration -.3327(4.08)
-.7991(3.84)
-.3493(2.03)
-.3126
(2.10)
---
---RESULTS
-.3148(4.20)
-.2515(3.63)
Experience .0097 .0070 --- .0166 .0101 .0084(2.55) (1.89) --- (2.16) --- (2.81) (2.55)
Specific Voc. Prep. .2893 .2606 --- . ...- .6594 .2982(2.45) (2.10) --- (8.08) (2.60)
STRATIFICATION MODULEUnion Member -- .3813 - .4705 -- 0 .3302
-- (2.58) --- (2.97) --- NOM 11111.111.M -- (2.40)
X Hinority--Industry --- -.7850 --- -.8673 --- .1 --- -1.0525--- (3.18) --- (2.14) (5.25)
Z Minority -- Occupation .4151111 , (2.16)
2 Black Male -- Occupation --- -8.0053 ---
-- (3./q) - -Z White Female-Occupation --- 1111.1111. -- -Z Black Female-Occupation --- MOWN* -INDUSTRY MODULEConcentration -- .9911 --- .8951 --- --- 1.1334
--- (5.56) -- (3.24) --- (7.02)
Union x Conc. -- -.7066 - -- -.7677
--- (2.77) --- --- (3.24)
After-tax 9rofit .0 --- 18.6527 - -- --- 6.3150- (4.61) --- --- (3.05)
Capital/Labor Ratio .0071 ---
(2.33) ---
Goiernment Demand 1 01.0
WORKING CONDITIONSMoDITEPhysical Demands WM* MI1111.
Negative Work Traits --- .1187 --- -- .1218--- (2.98) --- -- (3.23)
R2
.160 .252 .193 .424 .059 .277 --- .203 .336
SEE 1.0344 .9799 .9091 .7773 .8962 .7960 --- 1.0530 .9654
HEAN $3.50 $3.50 $2.59 $2.59 $2.36 $2.36 --- $3.29 $3.29
674 674 128 128 115 115 - - --- 820 820
192
a
175
White Males - All of the human capital variables with the
exception of insfitutional training are significant in the white male
equation. An additional year of education is worth almost $.16 an.
i
,
hour; in the south almost $.12. Non-migrants earn $-.33 less per hour
-.-----____ .
while each additional year of labor force experience is valued atd C q
almost $.01. Apparently experierde only begins to play a role in.1 -
earnings functions in the relatively skilled eccupations; within each
of the lower level,stratalthe wage-experience profile is flat. Finally
each unit in the SVP scale adds $.29 to,the hourly wage. Taken
_tegetherthese five factors explain_16 percent of the variance in
earnings.
Even within this relatively high skilled occupation stratum,
the stratification and industry_modules are reSpOnsible for a large._
f: -2 'increase in the coefficient of determination. The 4 rises to :252-
-after these variables are entered. The differential-in earnings due
o industry segregation -is almost $.25 an hour Which is eqdivalent1
--to_a 7.3 percent wage differential. There an Additional $.32
or 9.7 percent differential due to the crowdi4; of black men into
:-certain occupations = ( %BMOCC).
11 f ft
Union membership and concentration each increase the wage ratek
_
as well, but again the interaction term is negati Among. generally
k
competitive industries (MPF-4.20),_union membership adds $.'24 to-Y
hourly earnings or-7,4\percent. 'Among concentrated industries, however,
union membership apparently adds nothing extra to the wage rate.
_ Above a concentration ratio of .54, in fact, the statistical
193
,No_ union-
Union
176
Concentration (MPF)_
-20%'
$3:25 -0.65
$3.49 $3.60
-relationah*between unionization and_earningsis Slightly, negative,"
,In:the unorganized sector oligopolies pay an average of 12.3 percent
more than do competitive industries. In 'the union sector a 40 point
differential in the market pOwer factor is worth no more than an
additional 1L-cents or 3.2 percent. Yet, overall,. after we control fdr
human capital differences,ta unionized highly concenttated industry
'N\
pays wages which are $.35 an hour or 10.8 percent greater than an
industry which is unorganized and competitive. The importance of the
industry and stratification_variables thus prevails even among
relatively highly-skilled white ten.
One other point might be added. In the complete equation; the
negative working traits factor in the working conditions module is
The apparent negative effeCt of union membership on wage'rates in highly concentrated industries may-, in fact, have somesubstance to .it, It seems plausible that non -union concentrated=industries may pay higher straight-time wageS tri order to ward_off the
,sympathetic pressure for unionization. The unionized induserieS.on the
other hand may-settle on a total economic and rion:-economic_package_-:which-realizes a lower straight -time rate, but makes-up for it with_-_larger fringe benefits including_ longer 'vacations,'more numerous-_holidays, fully-paid medical insurance, life insurance, large pen-sip-rig,:
etc. In this case, unionization may mean lower straighttime wa3exates
but higher total remuneration. The regression cannot measure this -_
effect.t
i 194
4'
177
significant and of the expected positive sign. The average occupation
in this stratum for white men has_approximately one (.93) negative
, .---
'trait with a-standard deviation ofnearly one (.98).,-After controlling..'
for all other factors, a worker in an occupation with one extra
.
negative trait earns $.12 more per hour in compensatory payments.
What this appears to signify is that. amonerelativeliskilled (white
male)- workers, but not among the unskilled, wage rates/respond to
the '!quality",of the job in a compendatory manner..
Mack Males - An nsadditional year of education is worth-signi,-
Idantly 4.es6 to a black-male Worker than to a white male_in
=_= =stratum, What is more*-the-rate_of return:on schooling is--the_
lowest_of anrccupation group for black men refle!cting both.the-
higher oppOrtunity costs of additional education and the near constant
= /2-dollar value_ of schooling exhibited in,,each=of the strata. On the-
,
r hand, institutional- training pays'off handgomely, contributing
8 an hour to the wage rate. This most likely represents the
urn to specific training in fields like auto mechanics. As usual
ration is an important`mportant Iactor for black men, contributing hereft
9
21The differencelin meads test yields a e=2.57.
22The-dollar values and the internal rates of return for a year -_
schooling for black men by occupation stratum are:
ccu ation Stratum Dollar Value Internal R=Of R -
: 4733 -.0764
.0719n. a.
195
2.0<2.0<1.25
178
_almost $.35 an hour. The value of migration is similar to that found
inOCC STRATUM 1-3 end 5 although below that in stratum 6-9.,
Even within thaffrelatively skilled grOup,.the additi9n of theA w .
z
2-STRAT and IND modules more than doubles-theit to .424 and lowers _the_
.._
SEE_by $.13. Union membership is Worth $.47 an hour and there is no
apparent negative interaction between unionization and concentration
as there,is far white men. This may be due to the fact that on
,average black workers-in this stratum are found in much lesp-concen-
- .... . .-_
.
_trated industries. The mean MPF for white_men is'.46; that_for black-
16en only -.36. The regtession,_in this case, was not capable.-of
iso.,,ating the impact of concentration on the wage effect-of union
_membership_for the few black,workers_in oligopolistic industries.Il
"-Translated into percentage terms, the $,.47_wage increment dtig_torndion- ,,, .
affiliati iog issresponsible for a 19.5 percent wage differential;
6
--,
rate more than twice as large as thatfor:white men in the same stratum.
A
_ __-
This adds to the mounting evidence that unionization while important
for_white men is much more'So for blackmales in every single stratum.__
'eln this caste, access to the organized skilled crafts is the port of
entry to_higher earnings, human capital constant.
Industry segregation has an additional effect on relative
earnings, The ±laevaluation,of =MIND is valued at $.32 an hour..
In dollar terms, this is a bit larger than for any of the other-strata._
Tn the industry module, the same evaluation of concentration is valued
a6 $.54, indicating that the few
oligopolistic industries benefit
1.96
black men who do gain access to
.3°
substantially. Here "complex"
I -179
crowding seems to prevail. If strictly a'dditive, union membership,
DfININD, and concentration would furnish a wage differential of,,
$1.33, more than half the mean- wage for this group: This is, , ,
,;
equivalent to wage rates of $1,93 vs. $3.26 an hour.
White Females - Before. the introduction of the industry module111.
the only significant human, capital variable is schooling. Each _year
is worth 11.5 cents an hour, higher than in any previous group. After-v.-
the addition of the 'industry Module years of experience in the labor
force also appears as a significant factor in the earning generating
'function. EaCh year is valued 'at about 1.7 cents per houi which
_
ranslates into a wage differential of $.33 between a ilortan yho is
years of age and -one whof is 45..
_None of the- factors in the stratification in odule are significant_= 0
_including union membership; yet two of the induStry or "demand" side
variables are. According to theory there_ must be skther factors
-beclide unionization and minority segregation which serve to segment
A
the labor force. Imperfections in job information and inter- and
intra-labor market immobility due to geographical barriers may be
factors sufficipnt to offset the tendency toward wage equalization
as indicated by statistically significant variables in the industryse
These results, are, understandable in light of this stratum's
occupation composition. With 65 percent of the workforie as
"secretaries, only 10.percent of the workforce unionized, and -the mean
percent of white females in an occupation greater than 70 percent, the
)
187
t..
n
_
180
°majority of the variance idearnings, after pontrol/Ing for human
capital, is pr'bably due*to the pure institutional factor of "ability_, ,
to pay." Within the stratum, *Parity crowding and unionization are-
not particularly important, but which industry a secretary has access4
to apparently affects tie wage. A secretary in-an industry which has
a profit rate tlo above-tile mean will earn, on average, $.35 more per
. _
hour: In an industry which has ±]x more capital per.production worker
the average wage will be $.1T higher.23
The addition of just these
two variables is sufficient to more than quadruple the explained
variance and again suggests the tremendous importance of industry
attaqhment tor minority members of the labor force.!
'Black Females - SAMPLE SIZE TOO SMALL FOR STATISTICALLY
_ 4
Cross Race -Sex As in occupation stratum 6,9,_the individual
race-sex equations may underestimate the impact of.crowding. This is
particularly true because of the nearly 4mpfete segregation of white
women_in this stratum: The pooled regressions are therefore of value
n attempting to identify the true relationship between "crowding"
=and earninge;=
With the exception ot_institutionaltraining, all of the fattOrs
--in --the human capital module in the Pooled regression are highly
23It also seems plausible that "quasi-"siMpathetic pressure may
work within an industry. White collar personnel in indostries withstrongly unionized blue-collar wOrkforces may.benefit from the pro-duCtion workers' union without belonging. In the present_egyatio_n =it
is possible that, profits and capital-intensity are correlateewith theextent of blue collar unionism and consequently, produce this phenomenon.
188
1514
significant and explain about a fifth of the variance in earnings.
Unlike the weak effect of schooling in the previous stratum, here a
year of education is worth in excess of $.15,an hour. (In th& south,
an additional year of school is valued at 11.5 cents.) Migration.
adds over .$.-31 to av4rage hourly earnings and each year of labor.
t'force experience adds another $.01 0to the wage rate.
As expected, specific vocational preparation (SVP) is a potent
factor in the earnings function reflecting the,prime importance of
apprenticeship in the skilled trades. ,At the same time, the absence
of this factor,in-the equations for black men and white women and its
-Weaker presence in the white equation_ exposes the nature of:the_
link between demographic_characteristics and access -o occupations-_2
wh'ch-offer apprenticeship. The link runs first from race_and'sex_
upation and then from oCcupation to specific training. Access to
-fjob with an SVP _rating one -unit=higherthan-themean is worth___-_,
neatly $.66 an hour,- The addition of the stratification,,industry;._
and working conditions modules redUces the coefficient on _SVP by more_
-ban half and the further addition of the dummy for sex eliminateS
SW altogether._ Once again we have allowed this to occur because o
the- presumed causal relationship involved in the function.
The final complete equation including-variables for race'and
a
sex is shown in Equation (IV).
199
45
182
w'= 1.7244 + .1264 Schooling - .0354 School-South - .2601 Migration(7.52) (4.89) (3.84)
+ .0088 Experience + .-3991 Union - .8550' %MININD(2.98} (4.29)(2.66)
0- 7.3783 %BMOCC + .9492 Concentration - .8240 Union -Conc.
(4.55) 0 (6.10) (3.60)'
. .
+ 5.5602 After-tax Profit + .1304 Negative Work TraitsX2.78) (1v59)_
- _.8806 Sex
(8.28),: R2 = .370 SEE -= .9411
Except for SVP, the ineirity of the original human capital
equation_ia preServed-.
The dummy variable for_Tace is insignificant after controlling
_JorIMININD and black_male employment __(%BMOCC).. _141hile there may -
_still be some pure racial_disorimination within_industries, occupation,)-- =
and_specific_
rAilated more
AMoreover the
_firms,_the dominant stratification effect appears to be
'directly industrial and occupational crowding.
crowding hypothesis is supported by, evidence that the
Inclusion of the sex variable has only a minor deteriorating effect
oh-the coefficient_on MININD.and'none on %BMOC.C. Without the dummy
?the-±lo evaluations of these two factors are wroth $.37
:7ands $.23 respectively.. After the dumMy is added, the value on %MINIMA
falls by tinly 7 cents and the coefficient On %BMOCC actually rises
by $.11. Thus pure sex discrimination exists simultaneously with
crowding leaving the average female,$.88 an,-..hour worse off.
20 04
183
j On the demand side, union membership and concentration interact
in an almost identical fashion in the pooled regressidn as in the white
male equation. Among unorganized workers, a forty point difference
in the concentration ratio is responsible for a 12 percent difference
in average hourly earnings. In Competitive industries, union member-
=-ship is worth about $.23 or 7.4 percent; in concentrateCindustries,
unionization adds nothing to the wage rate. All in all there exists_ -
a 9.0 percent wage differential between similarly skilled workers in
unionized concentrated and unorganized competitive - indus- tries.
-union
Union
= Concentration-_--OTYY:-
-.207
+7.4%
_ Higher after-tax,profits also affect earnings in this stratum-_
independent of_unionization and concentration. The average wage rate
in an industry with after-tax profits ±la_higher=than the mean is-
.10 greater than for workers in the "average" industtf.c As in
white male eqbatison, t gative work traits is also significant-and--
the-expected positive sign, indicating again that at least at this
higher skill level, compensatory wage payments are necessary indUCe
-a-sufficient supply of labor to the more "distasteful" jobs.
What we,find_most interesting about .these results is that they
show that even' among relatively well-educated and skilled workers,
201
184
large wage differentials can be traced to factor's other-than differ-
ences in human capital and working conditions. This begins to
disappear only among the ver y most educated and skilled workers; those
--- in occupation-Stratum
OCCUPATION STRATUM 15-17
Unfortunately the number of blacks in tke,SEO's highest skilled
occupation stratum is too small to allow individual statistical
analysis---1Consequently the retalts refer only.to the white population
1
_-except for the pooledrace-sex-equations.
- -_i--,i-managerial occupations within this stratum he largest numbers are
found as accountants, insurance agenta,_draft Men, and secondary---...
White men are found in a plethora of sp cific profI essional and
*school teachers. Others are found.as.pharmacists and engineers. In
contrast over 68 percent of white women in this group are employed
, _,._-_-
in-just three occupations: aeAnimary school teachers, high dchool_
-teachers; and professional nurses.
.
The-white males in the sample average over 14 years of
schooling and nearly 45 percent have_had_some forM of institutional
training beyond their formal education Very few (t1%),,,are members of
- -trade_unions and the a-Verage_minori(ty_employment-im-their occupations_ . -
-ia-oniy 10.8 percent. The white females in this stratum have slightly
less'educption (13.0 years) andunly one-fifth have had any
vocational training outside of formal schooling. Union membership is
weak for women as it is for men (10%) but the average minority
0
202
185
employment in their occupations is three times .the male rate (34 %).
Of total minority employment in,this stratum over 90 percent'are White
women suggesting again the extremely small.proportion of-blacks.in
these occupations. Table 5.7 presents the regression results.a'
White Males - Among professionals and other highly skilled
..
personnel, formal education becomes the primary variable eAolaining. . %
_-.wag.-2 differentials. For white men, each .year of education is.worth1. 1
.
more than twice its value in any of the 'lesser skilled_strata. -An
- 1
additional year of schooling. is worth $.3i an hoUr,and there is AO--1. . _
- ,
_ differential associated with,where the scPolir3_was taken. _Only- in11\
7- -
'this higheit skikled4 stratum is there no discount for_-southern
education. The importance of formal schooling is, of course,,fully4
consistent with the type of-training usuallyrequired for these pto-
__-fessional_occupations.
Migration also plays a role. A $.48_wage differential exists:-
t i
between migianis_ and those whd4lave never moved -from the area in which,
1
they livecl_atp
age 16.' On a_full_ time full-year ,basis, this_is equivalent
,
almost a $1000 annual salary differential. Years of experience 4.p_
alsp especially.important_aading more than $.03 tO hourly earnings per
- -
year of labor force participation. Each year of eXperience translates
r *
-" an annual $60-salary premium. Finally specific vbcdtional\
preparation adds nearly $.70 per SVP unit to the houily wage. Altogether
-the human capital module explains about a quarter of he variance.
The addition of other modules co the equation doe not
appreciably improve the fit. The only significant variable is after-tax
203
o
186
TABLE 5.7
RECRESS/ON EQUATIONS:OCCUPATION STRATUM 15-17 BY RACE AND SEX
a
white Male Black Male
Constant -5.3929 -6.3036
: NUMAN CAPITAL MODULE ,
Schooling .3253 1.3192(7.26) ' (7.21)
Schoo:-Sbuth -Training -Migration -:4123*
(2.01) (1.84).
Experience .0318 .0351- (3.04) (3.37)
S pecific Vec. Prep. .67071 .6934(2.56) (2.68)
STRATIFICATION MODULE_---
===Unfon:Member-
XMinoritp-iIndustry?zt.- --
2-Minarity0ccupatioa1/11a-ckMale-Occupation-
X=Black Feiale--Occupation
-INDUSTRY-MODULE
Concentration
Unicti-x Conc.
--Af ter-tax Profit
CaPit'al/Labor Ratio
-
.0.10
SAMPLE TOO
SMALL FOR
SIGNIFICANT
RESULTS
NOON.,
4.1.04111,
15.5935 --(2.73)
--OW
--eGoverriment Demand -
NOPX2ING CONDITIONSMODULE .Physical DomanNegative Work Traits
2-Ft .247 .267
SEE 1.7809 1.7608
SAN , '$4.83 $4.83
287 287
*Percent Fectale-Oc cispation
White Female Black Female Cross Race-Sex
-.5661
.2620(3.59)
v..1.3813 .
.2853(4.13)
`N.,
SAMPLE TOO
\SMALL FOR
SIGNIFICANT
RESULTS
11.111 .10
-4.7987.-4-A604
.3513 .3196(8.34f (7.79)
-.4913 -.4622(2.20) (2.19.0279 .0320(2.79). (3.30).5322 .4544(2:18) (1.96)
0
--- 1.2463_=-- (222)_
-*7 - =2.47174(3.82)_
=0. 1=4110
. 0
.316 .422
.8954 .8385
$2.e3 $2.83
.7297(213)
14.6214'(2.80)
- 30 30
.204 ;
6
4.111411a .2541.7765 -1-.7110--
$4.63 $4.63_322
/
187.
profits; the tld evaluation is worth $.58 or 12 percent of the mean
wage,for this group. The addition of this factor increases the R2
by merely .02 and reduces the SEE by_only 2_cents. For all intents
and purposes stratification is not particularly responsible for the
variance in 1 earnings among white male professionals. 4' only among ,
1
this special group that the ('relatively) pure hums. , .2ital hypothesis....
holds. ' 6, ,.
-'1 White Females,-7 the.sample of white females is quite small,_but
sohof the vadance in earnings can be explained. In this case,
-
-however, ly education is significant in the huMan capital module
_end this_factor alone is responsible for 32 percent of the variance-in-
earnings". Each yearof formal schooling is worth at least $.26 an= _
k
-hour *hichtranslates into a sizable_9_percent-rate of return based)-
od- .method-ethod_ used throughout this analysis. This high gh rate'of return
s_no aubt,due to the effect of advanced_degrees opening up-access
o_occupations beyond teaching and nursing. Migration, experience,
-ti=
_
ene,0? do not' appear to explain any of the wage differential within
thii high skill strata ofswomen.\
_E- The inclusion bt he indusr ! module adds considerably to'the
explanation of earnings. The only,significant'factor isICOncentration,
--hut this-variable alone raises the coefficient of determination to _.422_
1 \,,
..
, -and-red uces the SEE from $.8954 to $'4-.8385. The ±10 evaluation of the=
_market power factor is valued at $.70 per hour which is equivalent
almost 25'percent of the mean wage for this group. Thus, while the
pure human capital model seems to account for'the overwhelming majority
205
,.188
of explained variance among highSkilled white men, other factors
still play a significant role in explAning the earnings of white
female professionals.
Cross Race-Sex - The cross race-sex equations contain, a small-
---
-number of black men and women,/}as well as whitee. The human capital
module results are similarto those for white men, but=in additionaz/
- 7-
variables in both the stratification and industry modules are significant.
The human capital-/mdaule contains schooling,-migration, experience,
and SVP., Each 'of the e!actors 'has a regression coefficient similar
-thosetin the white male equation. In addi,3; 1, before adding-the
b_ race and sex dummies the percent of female employment in an occupation
---_(%FMOCC) is highly significant and the ±lq test has_a value of4:76
__,anrhour. In the industry module both concentration and aftertaxi-_
___profits_are significant variable0 with the ±la evaluation Worth $.42
-add= _$.54 respectively. The inclusion of these three variables raises-
--the_R_.from .254 to'.315 and reduces the SEE from $1.7765 to $1.71314-
-When race and sex are added, only the dummyvariable for SexA.s,-
significant probably because of the verysmall number of blacks in-
-the subgroup sample. EqUation (V) gives these final reSults.-0
w = 4.9651 + .3120 Schooling - 4031 Migr tion + .0335 Experience
(7.72) (1.90) (3.53)
.(2-.14- .5524SVP - 1.4339 %FMOCC + 1513525
(2.99)
Profit
4) (2.05)
- 1.3944 Sex(3 8) RL 3 .338 SEE = 1.6814
206
,l89
_
The coefficients on the -human capital variables do not appreciably
change'after the addition of the dummy variable to the equation, but
the,coefficientbn %FMOCC falls from -2.47 to' -1.43 and concentration
. becomes insignificant. There'is obviously a large wagdiff ential
associated- with sex per se, yet the "crowding" factor still
significant as does after-tax profits. Given racial and.sexual differ-
-
ences in the labor force, stratification by occdPation and industry
plays some role in wage determination even at the top of the
occupational hierarchy. '
Again, further analysis of the regression results for individuaL
1=occupation-sfrata will be postponed until,the next chapter.
CROSS-OCCUPATION-STRATA _
The evidence-presented to this pant indicates that-Within-
;brbad occupation groups, stratiiiCation and industry variables
t
contribute to an-explanation of existing wage differentials,- In all
-cases these variables are of the proper sign, usually of large magnitude,_
= =- and -have relatively high t-values. Except in the case of the white
-.male equation in 0CC STRATUM-15-17, the addition oLAhe non-qiuman
capital modulles significantly boosts the coefficients Of determinetion
and reduces the standard errors of estimate. We can conclude that
within most occupation stratathe general model of wage'determination--
posited here is superior to any deyeloped in the tradition of pure
human capital or, for that matter, pure institutional theory; _
190
But the more severe test of the relative merits of human-,capital,
institutional, and stratification theory requires evidence across.-
occupational strata- As we have mentioned; it can well be argued that-
the findings-within strata do not ultimately test the theory since
individuals invest in human_capitar )5stensibly 'to move from one
, .
stratum to another. Testing the human capital theory within a single
stratum is therefore biased.in.favor of the institutional and stratifi-
cation hypotheses. This bias is eliminated by pooling the sample across7
-occupatibn-strata. The 'full impact of the human capital module can
% . _
then be measured. Table 5.8 provides these regression results. -
White Males - For the full- time -white male workfoice in the 1967 .
LEO sample, earnings averaged $3.42 an hour,with a standard deviation
2-$1.60---Based on eitheraisiMple R test or based on the changaln-
-the_standard error of the estimate, little additional variance appears-
to=be. explained by variables other than human-capital factors. The
complete equation including stratification, industry, and working=
=condition components increases the by by only .033 and reduces the--
:,.. .
_
---f--=_- --SEE_only slightly.
Each of the human capital factors is statistically significant
for white men with the exception of the vocational training variable:
-Each year of edu4ition is worth $.20 in hburly earnings if the
achooling was taken outside of the south._ Southern- education is valueid-1
at two cents less reflecting only a slight regional differential in
the returns to schooling for the white male workforce as a whole.
According to the rate of return methodology used throughout this study,
208
191
TABLE 5.8
. REGRESSION EQUATIONS:ALL OCCUPATION STRATA BY RACE AND SEX
White Male Black Hale
Constant
HUMAN:CAPITAL MODULESchooling
SChooI-South
Training
_Migration
-Experience
_ Specific VOc. Prep.
STRATIFICATION MODULE'
:Mnion-Member
--s,-
Z-Minothy--Industry
Z Minority-Occupation
Z-Blick:=Maleccupation
_2 white- -Female -- Occupation=
-2-_Black4Femaler-Occupation_:---
_-- INDUSTRY MODULE
:-:Conc-efitration
White Female 131adk. Female Cross- Race-Sex
.1911 .3353 1.5519 1.0838 1.5963 1.6647 - .9224 1.7209 .3113 1.0962 _
0
.2030 .1767 .0911 :.0.721
(15.62) (12.99)' (8.93) (8.05)
-:0199 -.0160 -.0471 -.0315 -.0249 -.0182 .0311 -.0215 -.0260 -.0186
(2.93) (2.37) (7.48) (5.73) (3.61) (2.91) (5:02), (3.77) (4;58) 13;44)
.0578 .0429 .0905 .0749 .1526 .1405.
(3.66) (2.95) (7.13) (7.64) 13.63) (12.66)
--- .2442 .2059--- (2.89) (2.83)
-.2884 -.2861 -.3745 -.2738
'(4.24) (4.28) (6.72) (5.65) ---. _
111
Immaew
.012? .0119 .0065 .0062 -...0095
(3.97) (3.90) (2.41) (2;03) (2.89)
--- .2315 .1620
(3.17) (2.35)
-.2230 -.104 -.3343 --_.30370.010 (3.66) (2.94) .(6.01) -(5.83)
.0059 ;0065.' (2:19) (2.50)
.1596 .1541 .0641 .0883 -.0478 -.0570 .0365*. .0468 -.2076 .1222--_
(8:67)- (8.42)- (3.93) (6.22)__12.00)-_(2.37)!--(1.66)_ (2.3) (13.22)A6:79)-
.;2460 OFR. .5452 *MY .2814 ---- .2850 --=--__:.3760f
, (2.00) (10.46) 1388) "=!.- (4.97)
-.5306 -.6746-_ --- -.5370 ---- -1;0032_ = _
(2.47)_
Uniiin:-x Conc.
-i Profit
f 1
CapitalfLabor Ratio
-_-Cowrnment Demand
-_--WORKING CONDITIONS.
---_-MODIME
...11
- Physical Demands---
Negativt Work Traits
--- 2- R
SEE
MEAN
.23
1.4142
$3.42
1850
(4.23)-- 13.00):-
.11 - 41111.0. a/MOM
.8599
(5.61)-
-.5671(2.49)
6.3688 8.2545 5.9787
(3.26)
4111.
-.1340-
(3.16)
.110-4.
"---;
(5.76)-
.0023(3.90)
1.0711(3.99)
flONIM4. =Mt
011,
12.96)
.0027
(3.02)
2.2590-(2.90)
-.232314.-79)
- -..256 :207. .421 .048 .155
1.3862 .8117 .6953 .9659 .9136
: $3.4/ $2.37 $2.37 $2.05 $2.05
1850 912 912 932 932
209
(6;59) (6-.31)
- ----(6.53)-
O 1.
--.3171 MY(2.64) (5.38)--=--==
=
6.5566
.001.0-
Maw.1
12-.00)1_ =
-(2.22)-_
-.1436 --o;
(3.27)
OaUPO. -4040
.291 ,
.564
$1.66,
397
.431 1 .238,- -.333-_
.5074 1.3270_1.2439 --
$1.66 $2.96-42.96__
397 2394 --2394
'192
4/7
the'average*white male worker reaps a 5-1/2 percent return by
remaining in school for an additional year (at the mean). This clearly
exceeds the rate of return earned by each of the minority groups'in*
the labor force; it is double the rate for black men and more than four
times the rate earned by white women.24
At least in relative terms,7,3
additional education is a good investment for white men, a finding
"consistent wi h virtually all human capital stusties.
Migration, experience, and on-theL-job training_ also play important.
./
parts in the wage determination process. Non-migrants earn, on average,
$.29 less per hour thamthose who have moved at _least fifty miles__1
a
-frOm7their place :residence at-age 16. This_Is_equivalentto_..:-
almost $600 per year for a ,full -time worker. Each year of labor force-_
.k,__
experience adds another 1.23 cents an_ hour to the wage rate. In t
annual terms this implies a $246 differential betwlem the earnings
of a fifteen year labor force veteran and a worker whoAlas been out
:Of school for only five years. -Finally each unit of specific'
_ Notational preparation is worth_$.:16per hour. Given_the full--range
_of this variable, there is a _$1.44 difference In_earnings_betweeda_
-worker in an occupation which requires only a short demonstration
-period and a worker whose occupation requires at least 10 years of on-_
--=:the-job apprenticeship. On an annual basis the impact of ,SVP has a
/24Theactual figures for the four race-sex groups are white =_
men -5.5%, black men 2.75%, white women 1.257,, and black women 4.75%:_
_These relative rates of return are consistent with apriori theory-and-re further explored in the text.
range of $2,880.25
While/the other exogenous
an-explanation of the variance,
are statistically significant.
-- _interact in thenOw familiar ma
1:93
factor modules add only slightly to-
six of the variables in these module&
_Union membership.and concentration
ner.
.
Concentration (MPF).
20%
$3._24
1.37 3.49
+4;-0%_ -2.8%
__These results indicate, thatunion_membership has only a marginal impact-- , _= :
_
-A-'4- --_7z-i- --
___on= relative wages in _both _competitive_and concentrated industries,_ a_
_-_---1
conclusion departing frommany _institutional 'analyses and_roughly
..:-
-:'= consistent with Weiss!su-results. In a_ similar' regression,. Weiss found-- '=- -=:---i-
--___ z , \thatunionization increased earnings -by at most 6=8 Pereentf$5ra-__
-- __
----comparable group of workers.2
---
6
, One example will serve-to indicate the magnitude of thepotential wage differential_based on these regression resulfs. A white
male high school drop-out with ten years 0 schooling_100uover migrated,
has worked in the labor force for five years and is pr'sently employed
--In an occupation which requires only a short tdemonstra on to learn its
basic skills will earn, on average, $2.15 an hour, Alternatively, a
.college graduate who has migrated, has 15 fears of laboA force e:xperi-andence and is presently in -a job requiring between one two years of
on-the-job training will earn $4.58 .pqN hour. This is equal to a $2.43
wage differential, the college graduate earning 113 percent more than
the high school drop-out-.= -,
26Weiss, op. cit., p. 108.
211
I
194
4
Concentration is, more important in the present analysis. Weiss
found a forty poineinCrease in concentration increased earnings by'
_only-3-5 _percent. Hereve-find,the- ncreade to be as large as 10.8
percent in,the non-union-sector. Again,we attribute our finding to
-the better measure of concentration used-in the present analysis. The. vr
weaker effect of concentration on earnings-among organized workers_
implies that unions in.the competitive industries haVe the ability to=y
Win wage contracts more in line with the pattern -set in the oligopoli- _
:stic sector while unorganized Workers in the competitive,sector do
===
not have_ t27
hisopportunity. 'Overall, a unionized-worker In a concen-
c-trated industry earns 7.7 percent_more than a sitilarly=akilled nOn-
union -worker in the classically-competifive sector of the economy._
Two other variables in the stratiic4tion and industry-_ modules
affect white male earnings. A ±la difference in MININD.is _valued
$.16 an hour while a similar ±1 evaluatiOn.,of after-tax profits
implies a $.22 differential. In ;both cases the effect_is statistically
eignifitant, but relatively minor be4qg only4.6'percent and 6.4 _percent_-
of_the man_ wage. In addition to these variables, the physical dem=
.,
factor hat a significant negative sign. Heavier work_apparentli earns --0
.,
I
implication we draw from these results is_thus,,,,at variance- :-
with the ov rall conclitsions of_Weiss. He Orites, "The implicafion=aeets--' _to __be that irms in concentrated industries do pay their _empioyees mo=re,
__-but that th get higher 'quality' labor In the bargain. Theincomes-won--_by union for their members more exteedwhatthose.wOrketa-
. _ ,-- -_
would earn i their_best alternative employments. "__ To the_extent_thait
_it_is possibl to differentiate between the effect ofconcentration:lnd
- -,-unionization, the present study appears to indicate that mOnopolyrepts
_ _arise more fro the produat market structure of an_industry than fro-A
te:present of-Unionization. Weiss,- p. 108..,. 0
0
.
a lowet,wage after controlling for human capital characteristics and,
industry attachment. Tbe coefficient on negative work traits is not
significantly different from zero. In neither case is there an%
indication of a compensatory earnings effect.
-Black Males - The regression results- for black males "are
tosharp contrast to those we have just seen. .'The human capital module
is responsible for less than half of the total explained variance in
earnings with the addition of the stratification and industry factors
reducing the standard error of the estimate considerably. The
essential structure of the earnings generating functions have a
cant __racial= component; as we shall see.
Every human capital factor in the black -regression is signifi=
cant. Schooling taken outside of the south adds $.09 per hour fbr
=every year completed. This is` less than half of the_ increment afforded
-comparable whites and amounts to a rate of return of-less than 3/
-1-
28 _..percent at the mean. Thi. low hourly increment and the in return-
are consistent with virtually all o f the studies that have b een made
of the impact of formal schooling on black maje,earnings.2 9
What is
more, southern schooling is worth only half as raucti as ichobling taken
elsewhere t presumably reflecting the poorer quality of southern black
28The difference in the coefficient on schooling betWeen the
white male and black male equation is significant at considerablybetter than the .01 level according to the difference in means test.e-r-16.77.
4'
29See_Hanoch, op. cit. and Bennett Harrisbft, op. cit. for two
Important studies in this regard.
213
196.
schools: The discount for southern schooling is much greater for
-black than for white men suggesting that the quality difference in
education between southern schools and all others may be primarily
race-related. In the-non-south, the'relative dollar return to
schooling between black and white men is (.0911/.2030)=.45.
south, the equivalent ratio is .24.
Vocational training is also a significant factor in the black
male earnings equation. This is the only group for whiCh this is true
In the
_implying that although institutional- manpower programs_ do not
appreciably affect the earnings of most workers, they do 1;euefit black
meit. Enrollment in a training_ program- is valued at $.24 an hour or
somewhat_ in Sxdess_of 10 percent of-the mean_wage._ Whether these
__programs actually__increase "endogenous" productivity_cannot_be direct.
meaSured,,of course. _What _ig ic'thesnifant coefficient suggestsmaycourse.
=V' _r
_-o4ly be that-black workers uto_have completed a_training program are-
=more likely to be hired for jobs_ that pay somewhat higher wages.= :
Migration is another powerful factor influencipg wages for_this _
_group, For the_black male workforce as_a whole, migration_is_worth
-0:_an_average of $,37jah hour, No doubt:much,of this_overall increment__
_
J
reflects the special beneficial effect of moving out of the south.4
The high rate of return attendent to southern emigration is most
-likely responsible for explaining the higher coefficient on "migration"
-compared to the parameter in the white male equation. Outside of the
south, migration may fail to pay off as handsomely for blacks as
214
197
it does for whites.30
Labor force experience increments earnings by $4065 an hour
per year. Each year in the labor force is consequently worth only
about half the rate for white men implying a much flatter age-earnings
profile. Finally each unit of on-the-job training (SVP) is worth
$.06 an hour. This figure too is less than half the coefficient for
white men. Part of this difference may be the result-of unspecified
non-linearities in the return to specific vocational preparation.
Alternatively, the smaller coefficient indiCates a real difference
the return to leach Unit_of SV_.
Taken together the_six human capital factors explain one-fift
_
-of the variance -in earnings among full-time black male workers. Thy- -, --:,=-_ -_
_addition_of the three remaining modules increases the coefficient-_ f
--_determination to .421. Union membership is extremely powerful_in_the
complete equation. The nearly $.55 wage differential_between union
_and_non-union work- ers represents an average_union wage which 25.7
_
percent greater than that received -by the average non-union worker.
ipusly exclusion from a_trade union has_a massive impact on the_ _
Evidence for this statement can bel-oundin Barry_ Bluestone,_William Murphy, and MarySIVenson,-Low- Wages and the Working Poor- _4
Unn.Arbor: Institute of- Labor and Industrial Relatiniversity,_ofMichigan-Wayne State University, 1973)=, p. 127. Regarding black males,_
-
-- "Mobility out of other-regions_of the nation (other than the_ -_ south) does not pay as handsoMely. Across-all-education_
-groups, moving out of the-Northe'abt is only slightly bene-ficial _fox those who move. to the North Central states or to
the West. All other moves actually increase the probabilityof pOor paying jobs."
215
earnings of a black male worker. While not a particularly important
factor for white men, unionization represents a most important route
to higher pay for the black male workforce. This is consistent with
both institutional and stratification hypotheses. The percent of
minority employment in an industry (%MININD) also affects,the_-
earnings distribution for this group of workers. A ±la difference
in XMININD is valued at $.22 an hour, Just slightly higher than the
effect on white male earnings.. I
The industry module in the filial equation has a structure-which___--
basically different from that-of white men. Neither concentration- - -
nor the interaction term are reported in the final equation, although
n_test runs concdntration (but not the interaction term) was
extremely significant and powerful. It was necessary to_drop concen7__-
tration from the final equation because its addition always destroyed
e_integrity of one of the human capital factors. All_of the
regressions which were prepared with concentration as one of the
_exogenous variables failed to include n experience It as a statistically
significant himan capital- factor. It was impossible to pin down the
reason for this deteriorating effect on the "expe'ience" coefficient.
As a substitute for concentration, other dustry variables
were significant in the complete equation with ut harming the_human_
capital module coefficients. These included the highly colinear after/
tax profit rate. The ±1Q evaluation of this variable is worth $.32
= an hour. Simp.ar ±1Q evaluations of the/capital/labor ratio and the
government expenditure variables are worth$.18 and $.20
216
199
respectively.31
Each of these effects takenIindependently have more
than a minor impact on the distribution of earnings. To the extent
that these effects are additive,-the industry module is quite powerful.
The case for "complex" crowding is convincing while the human capital
explanation leaves much;to be desired.
I
The actual importance of human capital in explaining'the -, \
existing wage differential between white and black men can be quanti-.
'fied by &sing the information generated in the regressions. The1
average wage for black male workers in 19_67 was ,$2.37 or 69 percent
of the average white male rate. The standard deviation was $.91.32-
\_
These three industry factors make perfect"quasi7"instrdmenta_
variables. They are colinear with concentration-but not_=with_ _
in-the human capital module. The partial (XtX) matrix for the trelevan,_
-- -factors- is reproduced-below. _ _
Partial (XtX)_Matrix=for Black_MaleaCross-Oecupation Equation
Concentration Schooling -Uperience-
after -tax Profit .5470 .-1464 -.0797:_ -
=Ratio .3706 --0614 .0758_ --
vement Demand .
1
Go rn 1352-1== *.0383 ..0069--_,
I
_ _
Variability-in-earnings in the sample white male population- -ig'nsiderably greater than in the black male group. The coefficient-of=Variatiol (V) for whites is .4678 whileIonly .3838 for the black,
'-f,
_male sample. Two factorsmight eXplailt.-this difference. One is that _
-the Underlying black male population _1. more homogeneous in human_-_capital and therefore more homogeneous in earnings. The other=-is- -that
the labor market treats black men as though they were more homogeneousin human capital than they really are (i.e. employers disregard human
--capital differences Or discount them). In the first case we wouldexpect to find a greater V forjlle-Ahuman capital characteristics of
,,: r_ a
A.or),A. 7
-200
If we subL.Zitute the black means into.the 'te equation the hourly
rate for black males rises to $2.58 or 75 percent of the white male
mean. Furthermore if we substitute black male means for the human
capital module, but, white male means'for.the other modules, the black
male wage rate increases to $2.73 ors80 percent of the WM average.
Assuming that SVP is a stratification variable'becduse it_is acquired
_ on the job after access to employment has been secured, the black male,
Wage now rises to $2.97 an hour or 87 percent of the white male mean.'
In this certainly plausible case factors other than human capital
account for over 56 percent of the BM/WM.differential and only_43-1
;
percent of the mean _wage difference between white and black males is
the white male-group.we find- the opposite to be true. Fot-each-of-the
jaitan_capital variables with theexception_of School-south,_thalr's-__Eor-white and black men are generally equal or the coefficient is
greater fot black men.ra
V .
bm-
Earnings
SchoolingSchool-SouthTraining-MigrationExperienceSVP
.4678 .3838 32.2188
.2773 - .4011 .6913
1.7467 ..8480 2.16171.6995 2.5823 .6581
1:2046 1.1601) .1.0383
14344 *.4220 1.0293
.3610 .4582 .7878
_This implies that the labor market is_leas sensitive to differences inendogenous productivity characteristics of the black male workforce.
-Larger relative variability in education, for instance, is notreflected in the variability in earnings. This does not necessarilyimply that individual employers who hire blacks totally overlook
ences in worker characteristics when choosing their_ employees. But-it.does provide another cogent piece of evidence-that the_labor marketlacing black workers is substantially restricted.
218
201
Atdue to measured differences in causally prior human capital yariables.
33
0
These results are summarized in Table 5.9.
TABLE 5.9
POTENTIAL WAGE RATES FOR BLACK MALE WORKERS UNDERVARYING ASSUMPTIONS ABOUT THEIR CHARACTERISTICS
Assum tions
(WMr-=$3.42)wm
Wa e BM/WM Ratio
MHC,Strat,-Ind, WC
--11-11111C (-SVP)
wmStrat, Ind, WC, SVP
$2.37
2.58
.69
2.73- .80
2.97 .87
BM = Black male estimating equation_WM = White male estimating equationbm = mean values for BM exogenous variableswm = mean values for WM exogenous variables
33The differential due,to measured human capital factors i&
calculated from:
b-MHC(-SVP)WM - WM--
HC
wm --Strat, Ind, WC, SVP
WM= - BM
219a
202
White Females - The overall structure of the complete white
.female equation is somewhat similar to that for the black male
workforce with two important exceptions. The first is that neither
the human capitalregresiion nor the complete equation are very gdbd
models of wage determination based on the coefficient of determination
or the standard error of the estimate. The second exception is that
the human capital,equation contains neither-training nor migration,,,both
of-which were significant variables inthe black male regression.
addition the sign on the experience variable is negative.
The unaugmented human capital equation explains only 4.8 percent
Of the variance in white female earnings and each'of the exogenous_
-variables is relatively weaL A year of_schooling in the non -south
=worth less than $.06 an,hour at the mean while a year of southern= =
schooling:is worth.only $.03. In the nonsouth this Is equivalent
to----a minuscule 1.25 percent rate of return on a year of educatioh, the
'lowest for any race-sex group. Based on this evidence, schooling ddea
not generally appear to be a very profitable investment for white
--women in terms of their own future earning:;. Vocational training is
not very profitable either. Although almost 11 percent of the sample
_had some form of institutional training, enrollment in such programs
does not have a significant-impact on earnings. As we have mentioned
previously; according to our regressions, oflly black men earn more due
to manpower programs. Migration plays no role either. This was not'
unexpected given the assumption that men migrate for economic reasons
while working women generally follow their husbands rather than seek
220
203
to maximize own earnings through geographical relocation.
When running the human capital variables alone, a negative sign
is found on the experience variable implying that more experienced
women earn less given equal years of schooling. As we noted earlier,
this result may be illusory because of measurement error. Given the
pattern-of fdmale labor force participation the "experience" variable
does not accurately measure the number of years in the labor force.
_However if human capital "depreciates" with mon-participation, it can
be expected that a woman who returns to the labor market after a period
of time out of the labor force will earn less than a woman who never
left work. This could explain a flat or negative earnings profile
with respect to the variable "experience" or, to be more accurate,
age. In the complete equation, the coefficient on "experience" is- not
significantly_ different from zero indicating a flat "experience-
earnings profile after controlling for all other measured factors.
Specific vocational preparation is barely significant at the--
.05 level. Each unit of SVP adds less than five cents to earnings,
an amount smaller than a third of that in the white male equation.
Again the relative size of the female coefficient may be biased
downward because of non-line;-:ity in the variable. But this seems
unlikely to explain such a large difference.34
34,Alternatively, the weaker earnings effect of on-the-job
training found in the white female equation may reflect a significantinteraction between this variable and other human capital factors.It can be hypothesized that each additional unit of SVP in combinationwith education or other human capital factors has a higher rate of
221
204
The addition of the three remaining modules increases the K2 to
.155 and reduces the SEE by $.06. As in the black male equation,
inclusion of concentration only came at the expense of violating the
proviso concerning the human capital module. Coefficients on both
schooling and SVP fell significantly when concentration was added to1
the equation. Consequently other industry variables were used as
quati-instruments.
Both union membership and MININD were significant in this
equation. Union membership is valued at $.28 an hour leaving organized
-workers earning 14.2 percent more than non-union employees. The dollar
,amount is approximately equal to that of white-male workers but wily.
-_about-half that of black men. in addition the ±la evaluation of
has'a valUe of $.22 an hour around a mean of $2.05. As in the black
male equation, after-tax profits, the ,capital /labor ratio, and Wet:,
government expenditures variable are all significant.
±la Evaluations
After-tax profit rate $.20
Capital/labor'ratio .19
Government demand .18
"Additive" Total $.57
return. Without some form of complementary investment, SVP alone is
worth little.Given the lower mean SVP for white women (SVP =5.28 vs.
SVPwf=4:16) this could explain the difference in the coefficients.
To test this we ran an interaction term including7GED and SVP and
another with schooling and SVP. Both variables were insignificant.
The lower coefficient in,the white female eqqation apparently either
represents the effect of specification error or implies a significantly
lower return to on-the-job training. The coefficients on SVP for the
white male and white iemale equations are significantly different at
better than the .01 level. (t' =3.96)
222
205
Finally the physical demands factor is significant, but once again
negative. Physically demanding'work is rewarded with lower wages,
other things equal.
Compared with the other race-sex groups, relatively less of the
variance in white female earnings is,explained' by the general model of
-wage determination. The use of interaction terms in the human capital
module might have improved the fit, but experimentation with these-
variables proved fruitless. Apparently there are numerous other0
factors not taken into account in the model which have special relevance
for whits women.3
35Conjecture-leads us ,to believe that one set of factors
determining earnings not taken into account in the general model relates
to the importance of earnings for women in various types of house-
holds_t__-Ceteris paribus, a Woman's earnings may-be inversely
relged to her family's ability to provide a sufficient income to
keep the family at a "satisfactory" or target standard of living. Where
the =woman's earnings are an important portion of the family's total
income, wt might expect more intensive job search by the female in
the household with earnings being the key argudent in her utility
function. Earnings may be a much less crucial factor in job choice'
in= families with sliffirient income from other household members or
alternative sources. In this case, two viomenvith equal endogenous
prOductivities may earn significantly different wages.Another set of factors that may be important in the earnings-
function for white women has to do with physical appearance and the
7 production of "psychological" benefits to employers. according to
/ Paddy Quick, women may be hired for other reasons than objectively
/ measured productivity; they supply their bosses (and their customers)
with a more or less pleasant social and psychological environment.-
The human capital characteristics measured in the present study may
not capture the-traiti which are "productie" in this respect. With
these factors missing, the general model fails,to account for a large
part of the variance in white female earnings. See Paddy Quick,
"Women's Work," Review of Radical Political Economics, Vol. 4, No. 3,
July 1972:
223
206
Although our equations leave a good deal of thewariance in..
earnings unexplained, we can, still estimate the impact ofILV)le human
capital module on the wage differential between white women and men.
This can be done as in the black male equation by varying assumption's
about-the mean values of the white female exogenous variables. The
results indicate that human capital is an extremely inadequate explana-
tion of the forty percent wage gap between white men and women.
.Plugging all of the white fetale means into the earnings
equation for white men increases the WF/WM ratio from .60 to .90. ,If
we use the white male means in the stratification, industry, and
---=t=Working conditions modules and the white fetalehuman capital4wans4---
thel ratio_ rises to .93. -Finally if we assume that- SVP is ,
-cation- factor rather than a lumen _capital variable and we, evaluate
- the_-white male -equation once_ again, we eliminate practically all-of
the difference in earnings between The two_ groups. Only .42/.40=5
__percent of the differential is directly due to sex=related-vdifferences
-In schooling, training, migration, and "experience." 'Given the
-Measurement of experienCe this may be a slight underestimate of the
*
full impact of human capital; but the thrust of the result still
stands even if we discount this variable by a large percentage. The
huge wage difference between white men and women cannot be attributed
to the latter's underinvestment in human capital. Crowding and other
forms of labor market discrimination play a much more critical role,
although other factors not inclUded in the model may be most
important.
224
207
TABLE 5:10 -
Aot
POTENTIAL WAGE RAGES FOR WHITE FEMALE WORKERSUNDER VARYING ASSUMPTIONS ABOUT THEIR .
CHARACTERISTICS
(WM = $3.42)wm.
Assumptions Wage BM/WM Rati
Strat, Ind, WC $2-.05 .60
WM--wf
HC Strat, Ind, WC 3.09 .90
-wfHC
Stra t, WC 3.18 .
==-HC(-SVP)
w`°- Strat, Ind, WC-, SVP 3.36====
Black Females - Black women are by far the poorest paid members
of the workfOrce. With an average wage Of $1.66 an hour in 1967,
black women earned only 48.5 percent of the average-wage for white
men and 81,percent of that for white women. Unlike white women, however,
the general model of wage determination_ is capable of explaining a
-good portion of the variance in their earnings. The human capital
.module alone is responsible for 29 percent of the variance and the
Complete equation has a corrected R2 of .431, _the'highest Among the
four race-sex groups.
Schooling plays a much more important role for black women
than it does for either of the other minority groups. This is primarily
225
208
due to the impact of edudation on occupationalmobility.36,
A year
of additional schooling (at the,mean) in 'the non7south'yields a wage_
increment of $.09 .°hour; in .the 'sduth, .06. This is more than0
\\
fifty percent higher than foi white women and, equal to the wage
increment for black men, Because of the extremely low opportunity costI
of additional schooling, the rate of return for black women is only_
secbnd to-that of. white men. A marginal year of schooling yields a._
4-3/4 percentrrate of return, 'only 3/4 of a percentage point behind
_ .
the white male rate.
Neither institutional training nor experience are significant in
thia--equation. But the coefficient on Migration iuggests-southern
--emigration is useful for black women-whether-main Motive for0
elocation is directly economic or not. A black woman who relocates
___ _ earns, on average,,P14,-2percent more,($.22) than _a similar Worker__
-- /_
co never moved more than 50 miles from her childhood home.
fl0
Spetific vocational preparation is not_ significant (even at the
-.05_ level) in-the human capital.equation. After Controlling for
industry Characteristiei and union membership, however, SVP becomes
0
significant at the :01 level with each unit of on-the-job training
yielding approximately the same return as for white women ($.047). .
The addition of the three remaining modules increases the R2 to
0
.43r and reduces the SEE to $.507. In dollar terms, union membership
36Fora more detailed Analysis of this poiut, see Bluestone,
Murphy, and Stevenson, op. cit.
p
2 26
A,
209
is worth $.29 an hour, an amount equivalent to that for both white
men and women. Because of lower average earnings, memberehip is
valued at 18,percent, more than four times the value for white men and
27 pei,eentmore than for white women._ Segmentation into minority-_
impacted industries isalso inch more iiportant for black women than
for any of the other groups. The±lo.evaluation of %MININD is-valued
--_at_$-.38 an hour, almost twice the_ effect found elsewhere. This is -_
fully consistent with other data which suggest_that black women have
historically been segregated into-a very small number of industries
and-occupations; many of which are related'to domestic and-personal
, _service. The one_ significant industry variable-in the final equation
s--Ooncentration; here the ±la evaluation is worth-$.16 an-hour.,
Together, union membership, -%MININD, and con_ ntration are-Worth_$:-.-83
an_shour, exactly half of the mean wage rate.
Evaluating the white male equation at the black female means
,_-urnishes added evidence of the qualitative difference in the earnings:
functions between the two groups. When the white male equation is
evaluated with all of the black female means, the wage, ratio rises
steeply from .485 tc .75. In this case the higher wages for black
women would be due to the higher gross returns on their human capital
and_the smaller impact of being assigned to minority dominated
industries. (See Table 5.11),
If black women were to gain access to the-same set of industries
as white men the wage ratio would rise still further to .84. In this
Case, human capital differences would be responsible for .16(1-.485)=31
210
percent of the total, wage differential. The other 69 percent would
be due to differences in the structure of the earnings functions
-(varying gross returns) and differential access to industries and
occupations. If we then assume that SVP is a stratification variable-,ii
//
the difference in human capital endowments is left to explain only1I6ri
, _. ,/
percent of the total wage differential. This isa-gpod deal mor
than for white women but substantially less than for black men.
TABLE 5.11
POTENTIAL WAGE-RATES FOR BLACK FEMALE WORKERS UNDER.VARYING ASSUMPTIONS ABOUT THEIR-CHARACTERISTICS /I
(WM==$3.42)Assumptions 'Wage
HC, Strat,'Ind, WC $1.66'
HC, Strat, Ind, WC 2.55
TitRC
wmStrat, Ind, WC 2.86
it
.75
.84
WMStrat, Ind, WC, SVP 3.14 .92
_All of the minority group results thus point, overwhelminLL;
to the importance of factors other than human capital in explaining
the large wage differentials between groups. Differences in schooling,
institutional training, migration, and experience can explain only
two- fifths of the differential between white and black males, only a
sixth of the BF/WM differential and only a twentieth of the differential
between white men and women. The remaining portion of the differential
ePriOfo4
211
is due to a combination of stratification mechanisms: unionism,
"crowding," and pure wage discrimination.
The relative unimportance= f h an capital differences may be due
in part to the specification of the pooled regresSions. The absence of
a log linear dependent variable, and interaction terms for education,
experience, and training may be responsible for this result. But other
investigations come to very similar conclusions as ours using different
.
techniques and.data sources. Blinder's study of wage discrimination
-uSingMichigen Survey- Research Center data cOncludesithat the amount of
Tintergroup wage differentials which can be explained by differences in
personal endowments is even smaller than that fdlund in the present _-
-37-study.- . For _the male_vage differential; Blinder_ concludes that Only
30-percent can be attributed to differences in endowments while virtu-
:01.ynone of the white male/female differential is due-to these
faCtors.38
Using -still different techniques-, both - Michelson and
-Siegel have also-questioned the importance of human capital endowments_;
in explaining white/black income differences.
:THE--1GRAND" POOLED REGRESSIONS-
--. The final three regressions reported.in this chapter are for
39
the total fulltime fullyear privately employed labor force. Even" if
37Alan S. Blinder, "Wage Discrimination: Reduced Form and
Structural Estimates" Journal of Human Resources, Fall 1973.
38Ibid., pp. 447, 449.
39See Stephan Michelson, Incomes of Racial Minorities (Washington:
The Brookings Institution, 1968) unpublished manuscript; and Paul Siegel,
"On the Cost of Being a Negro," Sociological Inquiry, Winter 1965.
74r7046 *a
212
somewhat imprecise due to their specifications, these equations do
clarify the dimensions of "crowding." The numerous caveats regarding
their interpretation have already been discussed.
When regressed alone the human capital module explains 24 percent
of the variance in all earnings. A year of schooling is worth $.15 an
hour ($.127 in the south) which translates into an average 4.5 percent
rate of return on the foregone income opportunity cost of schooling.
The training variable is.significant with enrollment in an institutional
vocational program worth over $.23 an hour.' Migration is worth $.33
;while each year of experience is*valued at nearly 6/10 of a cent and
:_-- each -unit of SVP adds $.21 an hour. The-mean wage for this 1967
composite sample is $2.96 with a standard deviation of $1.52.
Using this equation it is possible to estimate the range in
earnings under different assumptions about schooling; SVP, training, and
experience. For simplicity we assume throughout that schooling was-
taken outside the south("school-south"=0) and that migration had teen
undertaken ("migration"=0). These results are reported in Table 5:12a-c--
along with the estimated earnings for each of the individual race-sex
-Agroups calculated from the own occupation-pooled regressions. The
row W* in this table refers:to the estimated wage for the "grand" pooled
regression.- All of the estimates- are made from the human capital
_ equations reported in Table 5.8. The four rows below the dollar esti-
mates give the percentage differentials. from the grand pooled wage for
each of the race -seat groups. In all but a very few cases, white men
have-wages in excess of the grand pooled estimates while each-of the
o
230
213
TABLE 5.12a
ESTIMATED HOURLY' EARNINGS UNDER VARIOUS ASSUMPTIONSCONCERNING THE HUMAN CAPITAL. MODULE IN THE
OCCUPATION - POOLED, REGRESSIONS:
SCHOOL COMPLETED=8 YEARS
School
Completed 8 Years
SVP 2'
Training No Yes No Yes
Experience,5 20 5 20 5 20 5 20_-
$1.98 $2.07- $2.21 _$2.30 $2.81 $2:90 $3.04 $3,13 :
2.20 2.38 2.20 2.38 :2.84 3.-02 2.84 3.-02=_
2-.41 2.55= -2.65 -2.57 2.67- 2-.80' 2.-91
2.15 2.15 2.15 2.15 2.34 2.34 2.34 2.-34 __-
1.72 1.72 1.72 1.72 1.87 1.87 1-.87- 1.437
% '% x x % % %
1.1*-1.1._-)/W*-411.1 +15.0- --.5 +3.5 +1.1 +4.1 -6.6 -3.5_
ic7Vibm)/ii* 416.7 416.4_ +15.4 +11.3 -8.5 -7.9 -7.9 -7.0
W*-W')/W* 48.6 +1.9 -2.7 -6.5 -16.7 -19.3 -23.0.-25.2
'bfVW*-13.1 -16.9 -22.2 -25.2 -33.5 -35.5 -38.5 -40.3
231
214
TABLE 5.12b
ESTIMATED HOURLY EARNINGS UNDER VARIOUS ASSUMPTIONSCONCERNING THE HUMAN CAPITAL MODULE IN THE
OCCUPATION- POOLED REdESSIONS:SCI400L COMPLETED=12 YEARS
School-Completed 12 Years
SVP_ 2 6
Itaining No Yes No Yes
Experience 5
W* $2.59
3.01wm
bm
wf
2.67
2.38
2.08
(W*-Wwm
) /W* +16.2
(W*-Wbut)/W* +3.1
(W*41wf) /W*-8.1
(W*-Wbf)/1.1*
20 5 20 5 20 5
$2.68 $2.82 $2.91 $3.42 $3.51 $3.65
3.19 3.01 3.19 3.65 3.83 3.65
2.7/ '2.91 3.01 ' 2.93 3.03 3.16
2.38 2.38 2.38 2.57 2.57 2.57
2.08 2.08 2.08 2.23 2.23 2.23
% % X % %
+19.0 +6.7 +9.6 +6.7 +9.1 0.0-
+3.4. +3.2 +3.4 -14.3- -13,7 -13.4
1 -15.6 -18.2 -24.9 -26.8 -29.6
.-28.5 -34.8 -36.5 -38.9
20
$3.74
3.83
3.27
2.57
2.23
+2.4
-12.6
-31.3
'-40.4
215
TABLE 5.12c
A A
ESTIMAtED HOURLY EPRNINGS UNDER VARIOUS ASSUMPTIONSCONCERNING THE HUMAN CAPITAL MODULE IN THE
OCCUPATICN-POOLED REGRESSIONS:SCHOOL.COMPLETED=16 YEARS
SchoolCOmpleted
SVP 2
16 Years
6
Training No Yes No Yes ,
Experience 5 20 5 20 5 20 5 20
W* $3.20 $3.29 $3.43 $3.52 $4.03 $4.12 $4.26 $4.35
W 3.82 4.00 3.82 4.00, 4.46 4.64 ,4.46 4.64wm
Wbm
3.03 3.13 3.27 3.37 3.29 3.3,9 3.52 3.63
Wwf
2.61 2.61 2.61 2.61 2.80 2.80 2.80 2.80
Wbx 2_44 2.44 2.44 2.44 2.59 2.59 2.59 2.59
% % % %
(W*;Wwm)/W* +1.9.4 +21.6 +11.4 +13.6 +10.7 +12.6 +4.7 +6.7
(W*-W, )/W* -5.3 -4.9 -4.7 -4.3 -18.4 -17.7 -17.4 -16.6.,m
-18.4 -20.7 -23.9 -25.9 -30.5 -3,-0 -34.3 -35.6
(W*-Wbf)/W* -23:8- -25.8 -28.9 - -37.1 -39.2 -40.5
233
.
216
minority groups falls below the respective grand means. Black males
with 12 years or less of schooling and little on-the-job training
comprise the one major exception to this rule. ,There is also a.general
trend for the wages of minority groups to fall further behind W* as-the
amount of SVP, training, and experience increases. This trend is less
pronounced for increases in schooling. This all reflects the lower
' earnings elasticities (w.r.t. human capital) prevailing for minority
groups in the economy.40
40It is tempting to .interpret .1* in Table 5.12a-c as the
wage rates that would prevail for gives numan capital endowments in
the absence of "crowding." But this interpretation is not correct
except extremely restrictive assumptions. For W* to be theperfktly competitive ("uncrowded") wage, (1) the underlying distri-bution of, human capital must be identical for each of the subgroupsand (2) the ratio of the slopes of the sectoral demand curves mustbe inversely proportional to the employment ratio in the previously
segregated sectors. _ Proposition (1) is required in order for thegrand pooled regression estimates of W* to equal the weighted meanwage' estimates summed over the race-sex subgroups 6rs) . Proposition
/
(2) follows from the theory presented in Chapter III. The proof of
this is straight-forward.0 I-
Let (1) wi = al-biEl
(2) w2
= a2-b
2E2
and (3) w**= Wrs (w1yw2E2)/(E1 +E2)
with the first equality in (3) holding only if the human,-apital distributions are identical.
If wl = w2 in perfect competition, then from (1) and (2),
(4) El = (a17a2)/bi + (b2/b1)E2.
If a1= a
2equation (4) simplifies to the familiar inverse ratio'
(E1 /E2) = (b2/bi)
Substituting El = (b2/b1)E2 into (3) then yields
217
The addition of the three remaining modules boosts the toto
;333 vial a large number of significant variables. In the stratifica-
tion module both'%MININD and %MINOCC boast highly significant negative
coefficients. The usual ±la evaluations yield wage differentials of
$.38 and $.50 respectively. Union membership interacts with concentra-
tion to render the following effect:
No union
Union
Concentration
20% 60%.
I
$2.79,
$3.08
3.04 3.07
+9.0% 0.0%
In competitive industries, union members earn approximately 9 percent
more than workers who do7not-belong to a trade union, But in the
oligopolistic sector, union membership has no particular impact on '
'rlative. earnings. It seems reasonable to believe that the small ,
ffect reflects the relative extent of trade unionism in different
arts of the occupational hierarchy. With every few highly paid pro-
essionals and technicians in occupations with organized trade unions
as usually defined, the cross occupation union variable tends to
underestimate the impact of trade union membership in specific occupation
W* = f(b2/bi)wl+w21/[(b2/b1)+1]
This reduces to W* = w, = w2
in perfect competition.
Without explicit knowledge of labor demand in each sector it isimpossible to determine the wage impact of desegregation.
235.
218
and industry groups. Indeed, if who belong to professiodl1
organizations which behave like trade unons (e.g. the American
Medical Association) were assigned a duit9y value for membership it
It
seems likely that the impact of the union variable would be
greater in theseequations.
1
Concentration itself is relatively powerful in the non-unionA
sector, but again in unionized indusrries greater co centration does
not translate into additional higher earnings, yet, overall, a union
member in an olignoolistic ind7rry earns 10 percent more than an
unorganized worker in the'competitive sector/of the economy.
All of the other industry module variables are significant as
well. The ±la evaluations of after-taXprofits, the capital/labor
ratio, and government demand are worth $.23, $.13, and $.12 respectively.
Together they play a not insignificant role in explaining eAisting wage
differential7. even after controlling for the effect of concentration
and union membership. Finally, the physical demands factor is sigaifi-
cant but once more negative.
Adding the dummy variables for race and sex to this equation-pro-
duces some further insights. While most of the coefficients on the
human capital variables remain unaltered, the statistical integrity of
"training" is compromised no matter when the dummy variable for sex is
added. The simple correlation between sex and training is relatively
small (-.145), but apparently multi-collinearity between several
variables in the human capital module and sex is sufficient to produce
this result.41
No matter what the specific reason, however, there is
41 Investigation of step-wise regression results on the grand
219
enough other evidence to conclude that part of the explanation for
lower earnings among womer, is the result of less vocational training.
a
w = .8519 +0.1360 School - .0146 SchOol-South + .1010 Training*
(12.71) (2.81) (1.53)
- .2750 Migration + .0078 Experience + .1296 SW.
(5.50) , (3.12) (7.41)
- .3438 %MININD + .0423 %MINOCC* + .3161 Union Member
(2.16) (.28) (3.42)
+ .7738 Concentration - .6090 Union x Concentration
(6.03) (3.47)
+ 6.7310 After-tax profit rate + .0009 Capital' /Labor ratio
(4.09) (2.25)
+ 1.0652 Government Demand - .1389 Physical Demands
(2.71) z (3.72)
- .2920 Race --1.0648 Sex 1
(3.20) (14.29),2
R = .387 SEE 1t1922
Over a quarter of the white male workrs in the sample had some form
of institutional training during thei 'work careers. In contrast only
10.percent .of the white women in the sampseand 14 percent of
the black women reported institutional training.
Of even greater apparent interest, 'addition of the dummy - variables
severely reduces the, coefficient on the industry crowding variable
and totally eliminates the significance of the proxy for occupatio'al
=
segregetien. The coefficient on %MININD falls from -1.0032 to -.3438
pooled eqUation indicates that the multicollinearity apparently arises
between training, sex, and SVP. In an equation with just school and
sex, training has an F-value of 9.07 if entered as the next variable
in the regression. If training is to be added to an equation with
school, sex, and SVP, the F-value for training (if entered) falls to
2.80, well below the F required for statistical significance.
220
while its t-value drops from over 6.3 to less than 2.2. Meanwhile
the coefficient on %MINOCC declines from -.9075 with a t-statistic in
excess of 6.5 to +.0423 with a -paltry t of .28. At first glance
this suggests the near total absence of "crowding" after controlling
for "pure }discrimination."
Combined with other information, however, this conclusion seems
to be much more tenuous. Evidence from the (XtX) . mArix for the "grand"
pooled regression combined with the highly significant coefficients one
XCININD in virtually every one of the individual race-sex equations
strongly ,hint that (1) industry and occupational crowding is widespread
and that (2) workers ihminority-crowded industries are paid less
regardless of race-and sex.
TABLE 5.13
PARTIAL (XtX) MATRIX FOR "GRAND"POOLED REGRESSION
Race Sex %MININD /MINOCC
Race
Sex
%MININD
%MIN=
1.000 .006
1.000
.034
.470
1.000
.051
.637
.389
1.000
The complete elimination of %MINOCC from the cinal equation is
most likely'the result of collinearity with the better measured variable
for sex. In effect occupational "crowding" appears to be so complete
that it is impossible to independently measure its earnings effect.
While there is then no definitive prbof for
2a3
ontention that
221
"crowding" bears. much of the responsibility for the large wage
'differentials ound after controlling for human capital, the mass
of evidence points strongly in thisdirection. This concluion is. / .
reinforced by our,previous findings of a significant coefficient on
%MININD in a large majority of the individual equations, particularly
in'the lower occupation strata. Table 5,14 summarizes all :of the
%MININD results. If the bulk of these had been insignificant, we would
have been much more hesitant to conclude that crowdidg plays a critical
role in wage determination.
f
TABLE 5.14
SUMMARY OF SIGNIFICANT %MININD FACTORS(t- values in parentheses)
Race-Sex Group 1-3 5 6-9 12-14 15-17 Total
WM -2.0851 - .8005 - .7850 - .5306
(4.85) (2.91) (3..18) (2.47)
BM - .7213 - .5723 - .8673 na - .6746
(2:86) (2.34) (2.14) (4.23),
WF - .8593 -.5370'
(2.97) (3.00)
BF -1.4104 - .5404 na = lia' - 1.0994.
(4.21) '(1.96) (6.59)
Cross ..1.1829 -1.1728 -1.2192 -1.0525 -1.0032
(4.03) (6.22) (4.37) .(5.25) (6.31)
Cross w/R,S -1.0019 - .4138 - .7200 - .8550' - .3438
(3.51) (2.13) (2.90) (4.29)' (2.16)
Ort_I979
0
222
,Beyond this, the addition of the race and sex dummies to the
"grand pooled" regression strengthens the impact of the demand side
variables. Coefficients on concentration, after-tax profits, and
government demand increase after the dummies are added and the negative
coefficient'on the union-concentration interaction term declines.
Finally we note that addition of the race and sex variables
raises the to to .337, over 6d percent more than the coefficient of
determination for the human capital equation alone. Clearly then,
human capital is an important element in Wage determination for the
.whole labor force, but the story is much more complicated than all
that. This we shall see-even more clearly in the next chapter.
240
CHAPTER VI
AN EVALUATION OF THE REGRESSION RESULTS
In the previous chapter we reported all of the regression results'
and presented a preliminary analysis of each of the significant
variables. This analysis demonstrated the significance of institutional
and stratification factors in the determination of earnings and
provided substantial although not incontrovertible'evidence.of the0 -V
earnings impact,of inddstry and occupational segregation. The present
chapter extends this,analysis by estimating the overall magnitude of
earnings differentials for (a) workers who share the same human capital
but differ in industry and occupational attributes and (b) workers
who differ in human capital but.work in similar industries and
occupations. Instead of using an ad seriatum analysis of variables
as in the former.chaPter, the present evaluation considers the
variables in each module as a unit (or ad conjunctum). In this way
the combined impact of labor supply-restrictions can be measured as
well as the combined effect of the demand-side of the market. The
results confirm the significanceAf non-human capital factors for
virtually all members of the labor force and-especially for minorities
and all those on the lowers rungs of- the skill hierarchy. As in the
previous chapter, each race-sex group is separately analyzed concluding
with an investigation of the total labor force:
241
223
,The Methodology
There are a number of methods that could have seen used to
estimate the relative strength of human capital and no -human capital
' factors as determinants of personal earnings. A brief r view of some
of these and the reasons for discarding the traditional one serves
to introduce the multivariate method finally chosen for this eurpose.
The simplest method is probably an R2comparison or F-tes
Giv.en the nature of the regression procedure it is easy to measur
how much additional variance in earnings can be explained by the
inclusion of the stratification and industry variables. But it is
not' really the explained variance we are after. Instead we are
seeking an indication,of the size of- potential wage differentials
associated with the non-human capital factors. An R2
comparison or
F-test says nothing about this and therefore is-inappropriate.
The traditional elasticity measure used in most economic analysis
is somewhat more appropriate, but it too has a number of problems
' I
which cause us to reject it in this case. For one thing, point
elasticities may tell very little S)bout the relationship between a
particular pair of factors when evaluated at:points other than. the
mean. Constant elasticity measures can surely be calculated, but they
may bear little resemblance to the real, relationship between variables
when evaluated near the tails of the distribution. While this is a
relatively weak argument against the use of elasticity measures, -
there are additional arguments which are more cogent.
242
225
. It makes sense to compare price elasticities for various goods
or for various factors of production because the'unit of analysis
is the same throughout. But comparison of a "wage/concentration"
elasticity with "wage/profit" or "wage/education" elasticities does
not have the same appeal because of the very different units used to
measure the exogenous factors. Comparing the price elasticities of
apples and oranges has a common sense interpretation, but not so for
a comparison of the earnings effect of years of education and after-
tax profits.
A not unrelated problem arises m the non-marginal nature of
variation in the exogenous factors used in this study. Infinitesimally
small differences in human capital or industry and occupation
characteristics do not accurately characterize changes in these
variables. Normally we are interested in the effect of an additional
year of schooling--or even the attainment of a diploma or degree- -
not the.impact, say, of a one percent increase in schooling past the
eighth grade. The same can be said for concentration and other
industry factors. For thistreason, Weiss, for instance, uses given
levels of unionization and concentration in evaluating his equations,it
not elasticities.1 In the final analysis, what we are after is a
measure of some range of earnings over so ratlike of its determinants.
Such a range can be estimated by measuring continuous variables
at arbitrary distances from their means and measuring dichotomous
1See Weiss, "Concentration and,Labor Earnings," op. cit.
243
226
variables at zero and one (e.g. no-training/training). One con-
venient method is to evaluate individual variables at ± one standard
deviation from their mean 1.7lues as we did in-the last chapter.
For a normal dis -fbution this yields a range over the middle 2/3 of
the observations. For other than normal distributions, the range
seldom includes less than 1/2 or more than 2/3 of all the observations,
making this measure variable, but bounded. Such a measure, of course,
does not cover the full range of a variable's distribution and there-
fore in most cases provides a somewhat conservative estimate of the
potential total impact of a given exogenous factor:2
In our desire
to err on the conservative side if necessary, this is a satisfactory
measure if only a single variable is to be evaludted.
But by its nature such a single variable measure cannot provide
an unbiased estimate of the impact of a combination of factors
analyzed ad conjunctum. For present purposes such a technique is
required for ultimately we wish to estimate the earnings impact of
employment in a given multivariate "economic environment"--defined by
a combination of an industry's concentration, profitability, and say,
capital-intensity or the combined effect of industry and occupational
segregation. JThe ad seriatum measure tends in almost all instances
to give pward biased estimate of the combined range and in fact
may result in evaluation of the regression at points well outside of
2Recall Chapter V fn. 3 for an extended discussion of this
evaluation technique.
244
227
the data's regime. It may happen, for instance, that within all of
the observations in a given occupation stratum, no single individual
can be found in an industry which is simultaneously lo greater on each
of the separate industry measures. In this case it is obviously
improper to evaluate the equation by summing the ±10 wage differentials.
To overcome'this deficiency a multivariate measure was devised
that accounts for the actual variation in the exogenous variables,
taken as a unit.3
Use of this measure normally prevents an estimate
of a wage differential larger than the data's full regime and virtually
always smaller than the ad scriatum estimate. Consequently it tends
to further restrict the measured wage range due to industry and
occupation variables--once more yielding a conservative estimate of-
these factors. Separate unit estimates were made for the stratification
and industry modules. In evaluating the equations for the impact of
. "complex crowding," the two-estimates were then added together.4
The Z* Measure
The ad conjunctum measure used in this part of the analysis
involves estimating the standard deviation of a linear combination of
the continuous variables in a given module using the regression
31 am indebted to Prof. Malcolm Cohen of the Institute of Laborand Industrial Relations, University of Michigan for suggesting this
measure to me.'
4This may lead to a slight upward bias infthese estimates for
precisely the same reason that we rejected the ad seriatum measure, but
the_opposite signs on the stratification and industry module variablesprecluded the use of a joint ad conjunctum technique.
245
228
coefficients as scalars., The standard deviation thus derived will
be known as Z*, not to be confused with z-transformations or other
statistical parameters. A Z* range is calculated for eacliindustry and
stratification module based on the regression equations reported in the
last chapter. The derivation of this multivariate measure is generally
straight-forward.
Let aibe the estimated regression coefficient where X
ijis
the jth observation on the ith continuous variable. is then the
jth linear combination of the Xivectors.
+ a. X = Zm ml , 1
+ . . . -17 ,aniXm2 = Z2
a1Xln
+ a2X2n
+ + amXmn
= Zn
or in vector notation:
a1X1+ a2X
2+ +am X
m= Z
From this set of linear combinations, the mean of Zj (=2) can be
calculated as well as its standard deviation Z*.
_2)2
N-1
The measure ±Z* then provides a direct reading of the range in the
exogeneous variable due Ito the combined variation id the Xi's. In the
246
229
present cases ±Z)g is the ad` conjunctum measure for the effect of the
stratification module (excluding the dichotomous variable, "union
mem:.,er") while +1* is an analogous measure for the industry module.
Intuitively, --ZS is the wage differential associated with an
industry=occupation "envirunment" which has "dne standard deviation".
less minority emp' yment. The estimate +Zt is the wage differential
associated with a .dssive economic environment" assessed on the
basis of such factors as concentration, after-tax profits, capital-
intensity or governmeto- "Nmand.
The superio' y of this unit measure over the ad serlatum
technique cin be demonstrated, first, by specific example and then
more generally. It will be shown that the ad Seriatum estimate is
always biased upward except in .he improbable case of perfect
positive pairilise correlation between-exogenous variables. The
.ollowin6 simple but generalizable two-variable two-obserVationexample
demonstrates the bias in the ad'seriatum measure and the corrected '
estimate generated by the 1* method.
Assume a regression has been generated for Y containing two
5observations and two dummy independent variables,'X1
and X2'
In
order to simplify the example, let the final regression have the
form: Y = .251*
, + .25X2 ;Pc. With this limited information we can1,72
.
. ,L.
. ,
Compare the ad eriatum (Z') and aLiEDILECL2m (Z*) evaluations oft
the X module under the assumption of a positive correlation between
X1and X2. 1, this case, the values of the evaluation estimates
will be identical (Z'=Z*).
_5Obviously such a regression coold not actually be generated
- 40* -
Frf
230
Ad Seriatum (Z') Ad Conjunctum (Z*)
X1
X2
ok
.25 + .2511/ 1
Z' = .25a + .250.X1
= .1761+ .1767
Z' = .3535
X1
X2
Z
' [(.25 x 0) + (.25 x 0)) = -0
[(.25 x 1) + (.25 x 1)) = .50
Z* = .3535
In the opposite case where Xi and X2 are negatively correlated, the
Z' and Z* evaluations are no longer equal, the former generating a
value 119 different from the case positive correlar, but the
latter equal to zero:
Ad Seriatum (Z') Ad Conjunctum*(Z*)
Ai X2 X1 X2
( )
0 [(.25 x 0) + (.25 x 1)] .25
.25 .25[(.25 x 1) + (.25 x 0)] = :25
1 0
L' = .250 + .250X1
X2
= .1767'± .1767
Z' = .3535
because of its singularity.
248
Z* = .0000
O
a
231
In this case the "standard deviation" of the X module as
measured by Z*,is zero because of the perfect offsetting impacts of
X1
and X2
(given identical reeiession coefficients). This is, of
course, the correct estimate of the differential in Y due to the
combined effect of the Xi, for any t'gain" due to having the
Characteristic X1
is simultaneously offset by an identical "loss" in
Y due to the absence of X2,and vice-4 la. For the analysis at hand
this would be similaf to2situatior, where all industries with
greater than average concentration had less than average profitability.?
Empirically the zero-order correlations for the industry and stratifi-
cation variables are usually positive but far from unity. Consequently
the Z* measure corrects for potential overestimates generated by the
ad seriatum technique.
A more general demonstration of the properties of the Z* measure
can be, provided, again using two variabThs and two observations for
expositional simplicity. What is to be proven-4s that:
(6.1)
(6.2)
lim Z* = Z'
pX1X2'
++1
lim Z* F 0
PX X 4-11 2
if a1
= a2and where pis a zero-order correlation
coefficient between independentvariables
4
249
.(6.3)
232
Define the ad seriatum measure in the usual fashion:
Z' = alaX + a20X2
1
and let Z*.be the standard deviation of the linear combination of
independent vectors (Z). The.derivation of Z* is straightforward.
(6.4) . a1X11
+ a2X12
= Z1
aX + a2X = Z1 21 2 22 2
From (6.4) the mean of Z (Z) equals:
a1(X
11+ X
2) + a2(X12 + X
22) Z
1+ Z
2= = Z
2 2.
Therefore,
(6.5) a1X1 + a2X2 =
-411.-041°
The standard deviation of Z follows directly btdefinition:
az = z* = ozi - E)2,(14 -31/2
Solving for Z* in terms of X1 and X2 can be done by first solving
for the squared deviations.
(6.7) Z1-Z=a1 X
11+ a2X12
1a2R2
= al(X.. )X1)
a2 X12- X2)
250
233
Analogously,
.
(6.8) Z2- Z = a
1(X
21- X
1) + a
2(X22
- R2)
Then squaring both sides of (6.7) and (6.8) gives
2 al(' 11-n 251
2 - .2(6.9) (Z1-2) X X1) + 2a
1a2(X
11-1
1)(X
12-2 ) + a2(X12-X2)
and 2)2 = a2 (X121-=1'2 + 2a a (X 5iX) ) 1221-. ) (X22 2' 2 22 2
a2fx.
1
2
0
,And summing the squared delPations
(6.10) (Zi-i)2 = al[(X11 -X1)2 (X21-5(1)2]
+ 2a1a2[(X1lky(X12-R2)(X21-5i1)(X22-312)]
+ a2[~(X22 7(2) 2]
Finally dividing both sides by N-1 (=1) gives the variance in Z.
2(6.11) Z*2 = a2a
2+ 2a
1a2cov (X
1X2) a
2
2aX1
1 . 2
Now for Z*2
to equal Z'2, titer
2 2(6.12) a
1aXi
+ 2a1a2cov(X
1X2) + a
2a2
(a + a2aX
)2
X2
1 X1 2
2 2 2= a
1
2aX
+ 2a 1a2aX aX + a2aX2
1 2
251
which simply reduces to
(6.13)
234
cov(X1X2) = a
X1aXi
Finally, dividing both sides by ax ax provides a proof of (6.1).
1 2
(6.14) cov(X1 X2)=p x = 1
lr.2 7m.X1
2
ax1aX
2
Q:E.D.
To prove (6.2) divide both sides of (6.11) by and set'a1 =a2..a.
This gives
(6.15) 2 2 2 2a a
X1
2a2cov(X1X2)
a aX2
a aXi X2 X1 X2 X1 X2 X1 X2
Then setting cov(X1 X2)/(ax1 ax2) = P
(646)a a
1
a2aX27,*
2 X2a
2
2
ax1aX
2 2
aX
= --1 and cancelling yields
Remultiplying both sides of (6.16) by ax1 ax2 yields:
252
235
(6.17) 2 2 2 2 2 2Z* = a 0 - 2a aX
X2+ a ax
1 2 2
= (ao - au )2
1X2
Finally, taking the square root of both sides leaves an expression .
for Z*
(6.18)
Thus when P
Z* = aoX1
- ao =ra(o' - aX
)X2
X1 . 2
Z* ...: 0 if either of two conditions holds:
(1) a1=a
2=a=0
or (2) aX
= aX' when a1=a
21 2
Q.E.D.
The first condition is trivial, showing only that the X module has no
impact on Y when the regression coefficients on Xi
are insignificant.
Condition (2) is more substhitive, demonstrating that the impact of a
giVen module is zero when there is identical variance in all of the
. exogenous factors and the variables are inverse correlates of each
other. Thus the multivariate measure has the property of ranging from
zero to Z' as the correlation between paired explanatory variables
runs from negative to. positive one'. This is, of,course, a desirable
property for such a statistic.
253
236
The Results
In the actual estimates that follow, a Z* is calculated for
the industry and stratification modules wherever there are two or
more, continuous variables in a given module. Otherwise the equivalent
ad seriafum measure is used. Where the dichotomous variable, "union
member" is significant in a regression, it is evaluated at zero and
one and added linearly to the,estimate of Z*. Consistent with the
rationale for "simple" and "complex" crowding, the regression
equations are evaluated at (a) the mean for all variables (W), (b)0
then'at the mean for ail of the variables excluding the_ stratification
factors which are evaluated at ( ±Z * ±UN), and finally (c) at the mean
for all of the variables excluding those in the stratification and
industry modules both okwhich are evaluated according to the Z*
formula. This final statistic then measures the overall range in
earnings for a human capital constant population evaluatedin terms _
of'
±Z*S
±UN, and ±E*. All of these range or interval estimates are
based on the regression equations recorded in Chapter V. The tabular
results that follow report hourly and annual earnings intervals as
well as associated percentage differentials.6,7
Each race-sex group
6In terms of annual earnings, full -time full-year employmefit
is-assumed to be 52 weeks x 40 hours per week = 2080 hours/year.
7The two percentage earnings intervals are calculated in the
following way:
(-Z* +UN -IV) - (+X* --,UN'-Z*)
(1) "COMPLEX"(+X*
S-UN -E*)
2371
is reported separately and followed by the results for the labor force
as a whole.
White Males - As expected, the narrowest wage differentials due
toeXIsting-variation in non-human capital factors are found among
white men. Nevertheless these differentials are far from
inconsequential particularly in the lowest skill strata. '(See Table
6.1) The results for occupation group.1-3, for instance, establish a
perfect example of the "simple crowding" phenomenon. Holding human
capital fixed, a full $1.00 an hour wage differential is found based
on an evaluation of the STRAT factors alone. On an annual basis this
amounts to an almost $2100 interval around a mean of $5637. The worker
in a "permissive economic environment" (based on union membership
and the degree of minority crowding) can expect on average Ito earn
nearly 1-1/2 times (146%) the earnings of a similarly skilled non-union
worker in a minority-crowded industry and over 17 percent mere than
' the average wage in this stratum. In this particular case the
comparison is between a union,worker in an industry with 14 percent
minority employment and an equally skilled but unorganized employee ,
in an incluse?), which has over 46 percent of its labor force compo'sed
of white women and blacks of both sexes. In other strata the differ-
ential is by no means cc large, but still exists.
(-zg +um) - (4.4 -UN)
(2)
(-1-z* -UN)
255
"SIMPLE"
\
..
238
TABLE 6.1
WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM
WHITE MALES
OccupationStratum
1-3
+ UN + Z*
SUN
+I* - UN
+Z* - UN - Z*
5
-ZS + UN + Z*I
-Z*S+ UN
W
+Z*S
- UN
+Z*S
- UN - Z*I
6-9
-ZS + UN + ZI
-ZS + UN
W
+Z*S- UN
+Z*S- UN - ZI
Deviations from P
Total EarningsIntervala
W Annual W % I
$3.18 $6614 17.34% $1.00 45.87%
3.18 6614 17.34 1.00 45.87
2.71 5637
2.18 4534 -19.55
2.18 4534 -19.55
3.17 6594 10.45 .61 23.82
2.98 6198 3.83 .23 8.36
2.87 5970
2.75 5720 -3.83
2.56 5325 -10.45
3.41 7093 15.20 1.00 41.49
3.41 7093 15.20 1.00 41.49
2.96 6157
2.41' 5013 -18.58
2.41 5013 -18.58
56
239
TABLE 6.1 (Continued)
OccupationStratum
Deviations' from 171
Total EarningsIntervala
W Annual W
12-14
-Z* + UN + Z* $3.84 $7987 9.71% $ .85 28.42%
-2* + UN 3.75 7800 7.14 .44 13.29
3.50 7280 --
+Z*S- UN 3.31 6885 -5.42
+Z* - UN - Z* 2.99 6219 -14.57
15-17
-Z* + UN + Z* 5.11 10628 5.82 .56 12.40
-Z* + UN 5.11 10628 5.82 .56 12.40
P 4.83 10046 --
+2* - UN 4.54 9443 -5.85
+Z* - UN - Z* 4.54 9443 -5.85
All Strata
3.78 7862 10.52 .68 21.93Z-ZS + Ur +I
-ZS + UN 3.52 7322 2.92 .16 4.76
W 3.42 7114 --
+Z*S- UN 3.36 6989' -1.75
+Z*S- UN - ZI 3.10 6448 -9.35
I
The first row of statistics reports the interval between-Z* + UN + Z* and +Z* - UN - Z*
The second row reports the interval between -ZS + UN and +qc - UN
257
240
Occupation stratum 5, as one may recall, is comprised mostly of
operative and kindred workers. The evidence clearly indicates that
there is less variation in wages due to industry factors in these
fairly homogeneous occupations. Yet differences in .the extent of
minority employment by industry and concentration account fora $.61
earnings wedge between equivalent workers. On an annual basis this
amounts to a $1270 earnings gap or 24 percent. If we were to disregard
differences in industry demand characteristics and only evaluate the
regression for variance in the stratification module, the total wage
range would be only $.23 or 8.36'percent. Much of the tiotil wage
differential is consequently explained by differences in industrial
concentration given initial labor supply restrictions.
In occupation stratum 6-9, composed of many of the skilled
trades, union membership plays the critical role in the distribution
of earnings. Union membership alone is worth $.76 an hour (see
Chapter V) out of a total wage differential of $1.00, the remaining
gap due to the fact that apparently some white men are "trapped" in
occupations crowded with black male workers. The $1.00 an hour
amounts to a 41.5 percent earnings differential between workers of
apparently equal endogenous productivity. The difference on an annual
basis is $7093 vs. $5013.
A significant wage differential even prevails among white male
workers in the relatively highly skilled occupation stratum 12-14.
Here there is a $.44 earnings gap between workers who differ by ±Zt and
union affiliation and an additional $.41 due to differences in the
/4.58
241
industry module. Summed together this drives a 28 percent wedge
between the annual earnings in a "permissive" vs. "repressive"
economic environment.
Only for the very most skilled white male professional workers
is the differential relatively unimportant. Here non-human capital
factors are Iesponsible for nomore than a 12.5 percent wage gap
between similarly qualified workers and the full extent of this range
is apparently related solely to differences in industry profitability.
When we turn to evaluate the white male equation across all
occupation strata, 'thus accounting for the full effect of human capital,
we again find a relatively large wage differential due to stratification
and industry factors, particularly the latter. Stratification factors
(after controlling for the interaction between union membership and
concentration) produce only a 4.76 percent wage differential. Once
the Z* is added, however, the total earnings gap rises to $.68 or
nearly 22 percent. On an annual basis this amounts to a more than
$1400 differential, with earnings ranging from $3:10 an hour to $3.78.
While these industry and stratification associated wage differentials
are much smaller than for each of the minority groups, they are by
no means insignificant and certainly too large to ignore. The major
unanswered question is how to explain them.
Where much of the earnings differentials between race-sex
groups can be charged to discrimination in its many forms, this
explanation is mostly useless for the dominant white male group.
However, a number of possible alternative explanations can be4..
759
242
ascertained. One hypothesis consistent with radical stratification
theory maintains that wage differences among similarly qualified
white men are due to unspecified variation in the workers' social
class origins. Accordingly, higher wage workers have benefitted from
being nurtured in an environment of financially and socially well-to-do
families. Unfortunately we have not been able to control for this
factor due data limitations. Ultimately the social class hypothesis
may explain some of the wage difference associated with industry
and stratification factors, but at this point we have no proof.8
Another explanation might lie in compensatory wage payments
-which do not show 'up in the analysis of earnings or in fringe benefits
that are inversely correlated with straight-time hourly wages. What
evidence we have,on compensatory payments seems to indicate just the
opposite however. The phAical demands variable in the white male
cross strata equation is significant but negative. Little hard
evidence exists on the fringe benefit question, but casual observation
seems to indicate a probable positive correlation between wages and
8What evidence does exist on thisisubject tends to deny the .
importance of social class as a determinant-of the variance in income.In his study of Inequality, Christopher Jencks concludes that in factmost of the variation in men's incomes appears to be stochastic:
"Neither family background, cognitive skill, educationalattainment, nor occupational status explains much of the vari-ation in men's incomes. Indeed, when we compare men who areidentical in all these respects, we find only 12 to 15 percentless inequality than among random individuals."
A
Christopher Jencks, Inequality: A Reassessment of the Effect'of Familyand Schooling in America (New York: Harper & Row, 1972), p. 226.
"i%
ti
243'
non-wage supplements.
A more plausible hypotA'sis relies on,the existence of
widespread imperfections in information about job opportunitiO. This I
of course makes, a good of sense at least as an explanation of
short-run wlge differences. Such imperfections_could-well-explain--- --
wage intervals of the magnitude found in the higher skill categories.
Larger more permant_nt differentials, it would seem, require a more
cothplex hypothesis.
One such possible hypothesis can be derived from a synthesis of
thedries based bn tne work of Thurow and. Lucas9
(the "job competition"
thesis), 'Decker10
and Oi11
(the concept of labor as a "quasi- fixed"4
factor) and the institutionalists (the importance of "lock-i n" effects
in the supply of labor). ACcording to the job competition thesis,
,individuals compete for jobs based on their background characteristics,
not in terms of crape demands as standard neoclassical theofy suggests.
One can imagine a queue of jobs defined by a set of characteristics
the hourly, wage rate being o..e of the defining ?arameters.12
\9Lepler C. Thurow and Robert E.B. Lucas, The AmericariDistribution
,1 Income: A Structural\Problem, A Study Prepared for, the JointEconomic towmittee of the U.S. Congress (Washington, D.C.: U.S.Government Printing Office, March 171 1972), esp. pp. 19-39.
10Gary Becker, Human Capital, Chapter 11, CD. Ci.t.
11WQter0i, "Labor as a Quasi-Fixed Factor," Jce,Jrnal of'Political
. ,
Economics, December-1962.
12One very difficult q uestion is left unanswered by the job
comp etition model: what determtnes the distributiod,of wages in the
first place? If labor supply and demand facto:s are so weak as toloavl the wage indeterminate, what other factors define the actual wage
261Mln
Workers compete for these job/w43e slots by presenting themselves in
the job market to pote:ttial employers. Firms then choose employees
on the basis of expected training costs (given their background
characteristics), hiring first those with the lix..est expected employment
cost and then moving down the queue to higher cost labor if detaind
warrants.
If we apply this model over the business cycle, we can genetate
a pattern to that at any given point.in time workers of identical
endogenous productivity will be found in different job slots and thus
earn various wage rates. This will occur as\a worker who enters .he
job market during a period of tight demand will have a greatei
probabi/ity of finding a higher paying job while the worker who oins
the market in a contractionary period may have to acceptalwr
paying job for the same amount of search effort. If search c sts were
low, the fixed cost of hiring and training labor wen minimal, and
there were no substantial "lock-in" effects, earnings differentials'
would only be temporary for lower wage workers would con nually reenter
the job 'market in an attr-ot to gain employment in the gher wage job
slots consistent with their endogenous-productivity. A strong tendency
paid on a given job? One posstble answer is that s-pply and demand
are responsible for getting a wage,range"for every "job" but that
custom and inertia--as well as institutional factors including union
pressure - -are responsible for setting and holding the wage distribution
as it -is. Once established the pattern of wages changes only slowly
in response to real changes in supply and demand'. For the most part
the wage distribution is never in equilibrium accounting for a good
deal of-structural unemployment in all labor markets.
\245
toward equal returns for identical personal characteristics would
be-the consequence.
.In fact, however, labor is usually a "quasi-fixed" factor,
search *costs are often substantial, and "lock-in" effects are
extensive. Specific training costs will often be shared by both the
worker and the firm (with the shares depending on expected turnover
and quit rates).13
Once workers have invested in specific training
in a particular slot, their marginal products and therefore their
wages are presumably higher than in alternative employment. Consequently
a worker will tend to remain in a job for which he has already paid
for training rather than quit to begin a new job,at a lower wage rate
in hopes of working up to a higher one. Employers too will be
reluctant to dismiss already trained employees so as to hire replace-.
ments even if the potential recruits embody superior background
characteristics. Thus where labor has a high degree of "fixity," to
use ()its term, there will be a tendency for workers to stay where
they are (and employers to keep them) even in the face of fairly ,
substantial differences in hourly rates. This, of course, is fully
consistent with individual utility functions which posit that workers
attempt to maximize the expected value of lifetime income rather than
simply maximize their wage.
The foregoing eclectic theory, is obviously suggestive for the
more skilled workforce, those in our sample with high SVP levels for
13See Becker, Human Capital,,alcit., pp. 21-22.
Z63
246
instance. But the largest wage differentials, due to other than
human capital factors are found among the least Skilled workers,
presumably those. with. a low degree of "fixity." For them the
"quasi-fixed" factor theory does not directly apply, but an inst
tutional variant along the same theme does. Specific training/4d
hiring costs produce one form of "lock -in" effect, where the/More
common mechanisms are seniority privileges and non-vested p nsions,
both of which apply to the full occupation spectrum, the owest skill
strata included. In attempting to maximize expected lifetime incom1'
.a worker with many years of seniority and associated pensi n rights
will not move to a job.with a higher hourly wage rate if this means
sacrificing the employment security which goes along with seniority,.
(particularly in unionized firms) and the surrender of expected
retirement income. In this case fairly large wage differentials will
persist over time once the differentials exiit at all.
-Unfortunatelywe do not have any data tq test this',Ilypothesis,
but it seems a likely candidate to explain the substantial and
probably persistent wage differences found among all but the most
gilled white male workers. "Entrapment" through fixed training costs,
imperfections in information, and non-vested seniority and pension
privileges may very well be responsible for driving a wedge of as much
as $2100 in annual earnings between white male workers who have
substantially the same huMan capital attributes.
Black Males - Once we leave,the realm of white male workers,
the,impact of industry and stratification factors becomes much more
Z,64
247
significant. This can readily be seen in an evaluation of the black
male regressions. (See Table 6.2) In virtually every one of these,
there is extensive evidence of "complex crowding" with union member-
ship playing a consistently effective role in every stratum. The
percentage earnings gap is as high as 75 percent (0cc Stratum 1-3)
and the annual dollar difference, according to our evaluation
echni5ile; teaches almost $2800 (0cc Stratum 12-14).
Union-nlembership and Industry segregation are responsible for a
5 percent differential among black men in the lowest skilled occupa-
tion category. Adding the combined effect of di"ferences in concentra-
tion and government demand raises the total differential to 75.8 percent
or a $1.20 an hour range around a mean of only $2.17. The stratification
and industry modules apparently contribute about equal weight to the
overall wage gap. In occupation group 5 composed predominantly of
4 operatives and. janitors and sextons, the total earnings differential
is of almost identical magnitude (74.9%), but nearly wo-thir of
the total is due to stratification factors--mainly unio hership--
whil:e the remainder is due to the single industry factor, concentra-
, 4
tion (see Chapter V). This is in sharp contrast to the white male
regression for this stratum where we found only a small earnings
differential (23.8%). Of this only a quarter, was due to stratification
factors and union membership apparently played no role at all. The
rest of the relatively small $.61 differential was due to differences
in concentration..
ZE';5
248
TABLE 6.2
WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM
BLACK MALES
Occupation
"Stratum
Deviations from i4
Total EarningsInterval
W Annual W
1-3
-ZS + UN + ZI $2.78 $5782 28.11%"
$1.20 75.80%
-ZS + UN 2.51 5221 15.66 .65 34.94
W 2.17 4514°
+Z* - UN 1.86 3869 -14.28
+Z* - UN - Z* 1.58. 3286 -27.13
5
-ZS + UN + ZI 2.99 6219 25.10 1.28 74.85
-ZS + UN 2.75 5720 15.06 .80' 41.02
W 2.39 4971 --
+Z*S- UN 1.95 4056 -18.41
+Z*S- UN - ZI 1.71 3557 -28.45
6-9
-ZS + UN + ZI 2.87 5970 21.73 .95 49.23
-ZS + UN 2.70 5616 14.40 .60 28.57
ii 2.36 4909
+Z*S- UN 2.10 4368 -11.01
+zg - UN 7 Zt 1.92 3994 -18.43
2,66
1
249
TABLE 6.2 (Continued)
OccupationStratum
Total EarningsDeviazions from P Interval
W Annual 'W
.12-14
-ZS + UN + ZI $3.32 $6906 28.64% $1.33 66.83%
-ZS + UN 3.05 6344 17.76 .79 34.95
W 2.59 5387 7-
r $
-4-ZS - UN 2.26 4701 .712.74
4Z*S- UN - ZI 1.99 4139 -23.16
15-17
-Z* + UN + Z*
-Z* + UN /1/
SAMPLE SIZE TOO SMALL FOR SIGNIFICANT RESULTS
+Z* - UN
4Z* - UN:- Z*
All Strata .
-ZS + UN + ZI 3.01 6261 27.00 1.15 61.82
-ZS + UN 2.82 5866 18.98 .77 37.56
W 2.37 4930
+Z* - UN 2.05 4264 -13.50
+Z* - UN - Z* 1.86 3869 -21.51
250
The total wage gap in occupation stratum 6-9 is smaller than
in the other strata, a perplexing result at first glance. The full
interval is 49.2 percent, not much greater than the differential for
white men although still equiiialent to almost $2000 on an annual basis.
The relatively lower earnings gap is apparently related to weaker
effects of both unionization and concentration but even more so to the
virtual absence of any significant segregation factor. The perplexi
result is made comprehensible-once we recall that when segregation
in a particular race-sex group is overwhelming, the true earnings
differential may be empirically undetectable. The differential can
only be uncovered by evaluating the pooled race-sex regressions.
M6ving to the higher skilled occupation stratum 12-14, we find
the percentage earnings range among black men to be more than double
that of their white male counterparts and the dollar gap reaches a
maximum for any group in any'stratum ($1.33 an hour). About half
the total differential is associated with the stratification module
while the remaining half is due to differences in concentration. BaSed
on the evaluation procedure, estimated hourly wages for this group
span the interval $1.99 to $3.32. Unfortunately the data sample
does not provide enough observations on professional black men to
test whether the earnings differential substantially declines as for
white men.
In turning to an examination of the cross occupation regression,cx
one is immediately struck by the fact that the total percentage
earnings differential is almost three times that for white men. The
2 68
251
estimated range runs from $1.86 to $3:01'an hour compared with an
estimated range of $3.10 to $3.78 for white males. Of the full
$1.15 an hour wage spread due to stratification and industry factors,
$.77 is due to the "supply side" with the remaining amount the
effect of a linear combination of after-tax profits, capital/labor
ratios, and government demand. The average black man in the full -time
SE0 sample earned $4930 on an annual basis, but given "average" human
capital characteristics, the same worker could earn anywhere from an
estimated $3869 to $6261 depending on how fortunate he was in finding
employment in an industry characterized by a "permissive economic
environment."
Much of this massive earnings differential may be explained by
the same factors as we hypothesized for white men: compensating non-
wage supplements, imperfections in labor market information, and
lock-in or entrapmbnt effects. But in addition to these there is
considerable evidence of specifically race-linked segregation. The
estimated,STRAT module induced wage interval for the white male
pooled regression is only $.16 an hour compared with the $.77 range
estimated for black men. Part of this large difference is due to the
much stronger impact of union membership on wage differentials while
the remaining is due to the greater impact of the industry segregation
factor %MININT.
White Females - The tale told in the evaluation of the white
femae.regression results is a similar one, but even more difficult
to uncover because of a much greater degree of occupational
269
252
segregation. The estimated percentage differentials generally lie
between those of comparable white and black men. (See Table 6.3)
In the lowest skilled-category, only the stratification module is
significant but union membership as well as a linear combination of
both occupation and industry segregation provide a 42 percent earnings
interval with a dollar value of $.62 around a mean of $1.74. The
total differential in occupation stratum 5 is somewhat larger (54.6%),
but here the range seems to be better explained by differences in
industry characteristics wich the stratification module contributing
only $.24 to a total $.87 differential. Union membership is the only
significant STRAT factor in this regression.
Again as in the black male results, the earnings gap in occupation
group 6-9 is lower than in any other stratum (with the exception of
the professional group). The total gap is $.56 or 35.4 percent. A
smaller coefficient on the union memberhip parameter seems to suggest
the reason for this relatively narrow range in wages. But it is the
smaller variance in this factor due to the underlying high degree of
industry segregation that really explains this result.
This same.effect is nowhere more evident than in the top two
occupation categories where in both cases the regression coefficients
in the stratification module are insignificant thus yielding a manifest
earnings range of zero associated with these factors. The nearly 50
percent total wage differential in occupation stratum 12-14 appears
to be solely due to an ad conjunctum analysis of after-tax profits
and capital/labor ratios while the smaller 28 percent differential in
270
253
TABLE 6.3
WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM
OccupationStratum
1-3
-ZS + UN + ZI
-ZS + UN
ci
+Z*S- UN
+Z*S- UN - Z*
I
5
-Z* + UN + Z*
-ZS + UN
W
+Z*S
- UN
+Z*S
- UN - Z*I
-61-9
- ZS + UN + Z*I
- ZS + UN
i:i
+Z*S- UN
+Z*S- UN - Z*
I
WHITE FEMALES
Deviations from W
.Total Earnings
Interval
W Annual W % $
$2.11 $4389 21.222 $ .62 41.54%
2.11 4389 21.22 .62 41.54
1.74 3619
1.44 2995 -14.36
1.44 2995 -14.36
2.47 5138 22.69 .87 54.63
2.15 4472 6.96 .24 12.56
2.01 4181 --
1.91 3973 -5.00
1.59 3307 -20.65
2.14 4451 16.30 .56 35.44
2.03 4222 10.32 .34 20.11
14 3827
1.69 3515 -8.15
1.58 3286 -14.13
254
TABLE 6.3 (Continued)
OccupationStratum
Deviations from 14
Total EarningsInterval
W Annual W Z $
12-14
-ZS + UN + Z*I
$2.87 $5970 22.07% $ 05 49.63%
-ZS + UN 2.36 4909
W 2.36 4909
+Z* - UN 2.36 4909
+Z* - UN -.Z* 1.92 3994 -18.42
15-17
-Z* + UN + Z* 3.18 6614 12.35 .70 28.22i
-ZS + UN 2.83 5886 -- *MO.
II 2.83 5886 --
4Z*S
- UN 2.83 5886
41 - UN - ZI. 2.48 5158 -12.38
All Strata
-ZS + UN + Z*I
2.59 5387 26.01 .96 58.22
-.ZS + UN 2.36 4909 15.12 .49 26.20
W 7 2.05 4264 --
+Z*S
- UN 1.87 3890 -8.78
4Z*S
- UN - ZtI
1.63 3390 -20.36
272
255
the professionals category appears purely as the result of variance
in concentration. Labor supply imperfections not specified in the
regressions, such as those used to explain the white male wage
differential, are probably responsible for permitting the labor demand
variables to have such a significant impact on the estimated earnings
gap. Again it should be noted that the smallest wage interval is
among the professional class while large differentials permeate the
rest of the occupation strata.
In turning to the cross occupation estimates, we find a dotal
wage differential (in percentage terms) not significantly di ferent
from that of black men. In this case the total differenti is equal
to $.96 an hour or 58.2 percent.. A little less than half of this
differential is associated with the STRAT module while the remainder
is due, again as with black men, to a linear combination of after-
tax profits, eqltal/labor ratios, and government dematf1. The
overall wage interval runs from $1.63, just barely above the 1067
prevailing minimum wage, to a high of $2.59 an hour for women who
gain access to industries or occupations characterized by a
"permissive economic environment." In explaining these intra-group
diffeentials we might rely on the same hypotheses we posited for
white and black men and add the theory concerning different utility
,,functions for women in different objective situations that we outlined
in Chapter V.14
14See Chapter V, fn. 35, p. 205.
256
Black Females - The extraordinarily large wage differentials
found for black men are repeatedlrfob-latkwomen, u_th the exception
that in occupation stratum 6-9 the earnings gap is even larger.
(See Table 6.4) Being in a permissive economic environment can mean
as much as $1.17 improvement over those who are not as fortunate,
but given the very narrow range of opportunities for black women,
even a "permissive economic environment" leaves virtually all of the
' workers in the first three occupation strata with estimated annual
earnings below $5,000./ In each of these cases, the largest part of
the overall differential is due to stratification factors with union
membership significant in every regression.
/
An evaluation of the pooled strata equatim turna_pufto yield
an ear..ings range which is almost identical in percentage terms to
those found for the other two minority groups, although in this case
a greater pibpoktion of the total differential is associated with the
stratification factors. Only concentration,is significant in the
industry module and at best variance in this measure adds $.15 to
the $.82 differential. ,Table 6.5 demonstrates the near identical
percentage differentials for the three minority groups. This striking
similarity in the overall earnings differential is in sharp contrast
to the much smaller interval associated with differences in industry
and stratification factors for white men. Clearly the minority
guitips have something in common which they do not share with the
-dominant group in the labor force and it is far from their advantage.
274
257
TABLE 6.4
WAGE INTERVALS DUE TO STRATriICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM
. BLACK FEMALES'
OccupationStratum
\Deviations from W
Total Earnings'Intervals
W. Annual W
1-3
$1.79,
1.71
1.36
1.13
1.04
, 2.21
1.86
1.53
1.39
2.36
2.17
1.72
. 1.37
1.19
$3723
3557
2829
2350
2163
490c
4597
3869
3182
2891
4909
4514
3578
2850
2475
31.95%
25.73
-16.91
-23.50
26.88
18.81
--
-17.74
--25.26
37.20
26.16
--
-20.34
-30.81
$ .75
.58
.97
.68
1.17
.80
72.47%
51.32
69.78
44.44
98.31_
58.39
\\N-...
-2* + UN + Z*
-2* + UN
W
42*S
- UN
+Z* - UN - ZIS -
5
-Z* +'UN + Z1S L
-q + UN
W
4Z*S- UN
42* - UN - Z*I
6-9
-ZS + UN + 2*I
-ZS + UN
'T/
+Z* - UNS
4-2*S- Ui- Z*
I
275
258
TABLE 6.4 (Continued)
Total Earnings
Occupation Deviations from W Intervals
Stratum W Annual W
12-14
-2* + UN +Z*
-Z*S+ UN
SAMPLE, SIZE TOO SMALL FOR SIGNIFICANT RESULTS'
+Z* - UN
+Z* - UN - Z*\ I
15-17
-Z* + UN + Z*
-Z* + UN
SAMPLE SIZE TOO SMALL FOR SIGNIFICANT RESULTS
+Z* - UN
+Z* - UN - Z*
All Strata
-Z + UN + 74 $2.12 $4410 27.71% $ .82 63.07%
-Z*S+ UN 2.05 4264 23.49 .67 48.55
,.'
W 1.66 3453
+Z*S- UN 1:38 2870
+Z* - UN - Z* 1.30
-16.86
2704 -21.68
Z76
259
TABLE 6.5
POOLED OCCUPATION REGRrSSION WAGEINTERVAL ESTIMATES
or
Race-Sex GroupDo ar
Dif rential
PercentageDifferential
Black Males $1.15 61.82%
'White Females 96' 58.22
Black Females .62 63.07
White Males .68 21.93
f
Cross Race-Sex - The individual race-sex equations mask the
of "crowding" as the extent of segregation rises beyond some
point\ Nowhere is this more true than among higher-skilled white
females where occupational segregation is so extensive that the measured
effect of the stratification module is zero. For this reason the
pooled race-sex equations must be evaluated to correctly estimate
the impact of the industry and stratification factors. The results
/con;:irm a significant earnings effect in every occupation stratum ana
for the labor force,ias a whOle. (See Table 6.6)
This effect .is by far the greatest in the low skilled occupa-
tions. The total estimated earnings range in occupation stratum 1-3
is a startling $3350 around an annual full-time mean of $4722.
Workers in a "permissive economic environment" earn 35.percent more
tnan the average wage for this group and more than twice (110%) the
wage earned%by those in overcrowded unorganized competitive industries.P
Minority segregation by industry and occupation, com)ined with union
260
TABLE 6.6
WAGE INTERVALS DUE TO STRATIFICATION ANDINDUSTRY FACTORS, BY OCCUPATION STRATUM
ALL RACE-SEX GROUPS
'OccupationStratum
Deviations from WTotal EarningsInterval
W Annual W
1-3
-ZS + UN + ZI $3.07 $6386 35.24% $1.611 110.27%
-ZS + UN 2.97 6178 30.83 1.41 90.38
T',/ 2.27 4722
+Z*S- TN .,
1.56 3245 -31.27
+Z*S- UN - ZI 1.46 ' 3037 -35.68
5.
+ UN F Z* 3.08 .640 24.19 1.21 64.70
-Z* + UN '2.85 5928 14.91 .75 35.71
2.48 5158
+Z* - UN 2.10 4368 -0.32 /
+Z* - UN Z* 1.87 3890 -24.59
6-9
-ZS + UN + Z*I ..
3.36 6988 29.23 1.54 84.61
-Zt + UN0
'1.36- 6988 29.23 1.54 84.61
1,71 /.6C 5408
+Z*S
-- UN 1.82 3786 -30.00
+Z*S- UN 2*
I1.82. 3786 1*-- -30.00 ..
261
TABLE 6.6 (Continued)
OccupationStratum
,Deviations from T.1
Total EarningsInterval
W Antival W
12-14
$3.90
\ 3.58
3.29
3.08
2.77
5.35
5.01
4.63
4.2
'3.92
3:67
3.42
2.96
2.56
2.30
:
''
$8112
7446
6843
6406
5762
11128
10420
.9.630-,
8840
- 8154
7634
7114
6157
5325.
4784
18.54%
.8.81
°
-6.38
-15.80
15:55
8.20
-- -
-8.20
-15.1 55
23.98
15.54
-13.51
-22.29
$1.13
.50
1.43
.76 .
1.37
.86'
:
40.79%
16.23
36.47
17.88, ...
59.56
33.59
-Z* + UN + Z*
-ZS + UN
W .v.
+Z* - UN ,
S
+2*S
- UN - ZtL
15-17
-ZS + UN + Z*1
-ZS UNS__W
+Z* - UNS
+Z*S- UN - ZI
All Strata
-ZS UN + Z*S I
-ZS + UN
itt
+Z* UN
+Zt - UN - Zt
Z79
262.
membership, is responsible for a $1.41 earnings differential while
concentration adds another twenty cents to the overall range. In
annual terms the wage interval runs from $3037 to $6386 with human
Capital evaluated at the occ group means. overall 110.27 percent`
wage interval compares with 72-76 percent intervals for black males
and females and 42-46 percent for white men and women suggesting a
.strong racial and sexual component in the industry and occupation
distribution of low-skilled workers.
In cccupation stratum 5 this "discrimination" component appears
less pronounced as the pooled percentage earnings differential falls
within the range of the separate estimates for each race-sex group.
Overall there is a $1.21 wage interval around a mean of 4.48.' A
little more than half (35.7%) of the total interval (64.7%) is produced
by the STRAT module while the remaining is due to a linear combination
of concentration, after-tax profits, and capital/labor ratios. On an
annual basis the estimated interval is more than $2500 running from
$3890 to $6406.
Turning to occupation stratum 6-9 we once again find an
indication of the massive effect of industry and occupational discrimi-
nation. In none of the individual race-sex equations were any of the
industry and occupation segregation variables significant (with the
exception of union membership). But in the pooled regression three
of these factors are significant and powerful. Analyzed ad conjunctum,
%HININD, %MINOCC, and %BMOCC plus union membership are resnonsAde
for an 85 percent earnings differential. Of this total range,
263
unionization is responsible for a little less than half ($.67) while
the other three crowding variables make up the remainder of the $1.54
interval. After controlling for these factors, differences in
industry structure have no additional effect on the wage range
suggesting "simple" but substantial crowding.
In the two higher skilled strata as well there is evidence of
sizable wage differences associated with the industry and stratification
factors. There is a $1.13 wage gap (41%)'in occupation group 12-14
with an interval of $.50 associated with union membership combined
with a Zoc evaluation of MININD and %MINOCC. The remaining $.63 is
due to a linear combination of two industry variables: concentration
and after-tax profits. Even among professionals there is a 36.5
percent differential or an almost $3,000 annual salary interval after
'controlling for human capital characteristics.15 This is primarily
due to the sex-liAked segmentation of the professional labor market.
About half of the interval in this stratum is due to the single
stratification variable %FMOCC while the remaining amount is associated
once arain with a linear combination Of concentration and after-tax
profit:4.
15As has been the case throughout, we have evaluated these
equations assuming that SVP is a true human ,..apital ce;oponent, not a
function of Industry'or occupation segregation. Of course if we were
to.interpret SVP as a stratification variable--for which there is a
good deal of ity.tification--the wage interval would be muct larger
in a number of these equations including tha present one. ,ater when
we evaluate the effect of human capital, we shall assigrPthe whole
weight of the SVP factor to this module, surely an overestimate of
the pure human capital effect,
264
Finally we come to the "grand pooled" regression for the
whole labor force. Here we find for a human capital constant popula-
tion a total range of $1.37 an hour or $2,850 a year on a full-time
basis. This amounts to a 60 percent earnings differential between
workers in a!"permissive economic cnviponment" and those, who for'one
reason q another, are consigned to industries which are on the
"periphery" of the American industrial structure--industries which are
non-unionized; impacted with minority groups, low.profit, labor
intensive, competitive, and lacking support in the form of go ernment
contracts.16 A little more than half (33.6%) of the total interval
is associated with labor supply restrictions *bile the rest is due
to differences in industrial characteristics.17
Such large differences
16See Robert Averitt, The Dual Economy: The Dynamics of American
Industry' Structure (New York: W.W. Norton & Co., 1968), esp. Ch. 5.
17We should emphaS'ize again that these are conservative estinates
b'cause of our evaluation technique. If we were to estimate the wage
interval over the total range of the exogenous variables rather than
at ±lo around their' means, or if we were to use an ad seriatum measure.
of the interval, we would find a much larger earnings range due to
the'industry and stratification factors. Evaluating the "grand'pooled"
regression ad :seriatum rather than ad conjunctum increases the total
wage interval co $1.71 and the percentage differential to 83.5 percent.
Instead of an annual income spread estimated at ;72,850, the ad seriatum
interval is $3,550, twenty-five percent larger. The correlation
matrices for 'lie relevant varlet-les indicate the reason for the lower
ad sonhInctum estimate.The zero-order correlation between %MIWIND and %MINOCC in the
stratification module is .3887. Tne industry module correlation matrix
has the follouing values:
Concentrtn. At,:TX Pr. K/L Ratio Gov't Demand
ConcentraticnAfter-tax profitsHaGov't Demand
1.0000 .33201.0000
, .4531-.08721.0000
.0997-.0092.0533
1.0000
mmlim
due to factors other than measured human capital surely calls into
question Leonard Weiss's conclusion--and the assumption of most human
capital theorists- -that "The general picture is one of fairly
efficiently working labor markets, even where substantial monopoly may
exist."18
What we have found in this extensive analysis is significant
evidence of widespread mismatching between endogenous productivities
and marginal products. Workers with substantially the same human
capital attributes earn substantially different wages, much of this
apparently related to industry and occupation "crowding" with
variations in industrial structure and performance adding to the
overall Wage diSpersion. The personal earnings distribution, we have
shown, is to a far-reaching extent a function of institutional
factors well beyond the purview, let alone control, of the individual
worker.
The Relative Impact of Human Capital andNon-human Capital Factors
Before bringing this analysis to a close, there is one
additional question that warrants our attention. We have estimated
the earnings differentials associated with industry and stratification
factors, but not those which are due to variation in human capital.
How large are these in absolute terms and relat::.ve to the size of the
Z*Sand Z* intervals?
To evaluate the human capital variables, we have resorted to an
ad seriatup measure so as to avoid as much as possible the potential
18Leonard Weiss, 2p. cit., p. 116.
',Z83
266
error of underestimating the full impact of this module, again if
anything biasing our overall estimates in favor of the human capital
hypotheses. Each of the continuous human capital factors (schooling,
experience, and SVP) were evaluated at -Ho around their means while
the dichotomous variables (migration and training) were evaluated at
zero and one.19
While the human capital factors were allowed to
vary in this Way, the values for the stratification, industry, and
working conditions variables were set at their respective means. TwO
ad seriatum estimates were made: one for differences in s'..hooling
alone (ED-interval) and one for the 'complete human capital module
(11C-interval). These were then compared with the earnings differ-
entials associated with the industry and stratification factors
P -
(2- interval) by computing the ratio of the 2-interval to each .of the
human capital ranges. The final numbers that result have no cardinal
meaning, but can be compared in ordinal fashion. The results are
found in Table 6.7,
The findings for white men are especially interesting. Although
the ranking of the o.cup.ation strata is imperfect because of overlapping
SVP scores, there still is a general ordinal trend in the skill
colitent of jobs as one moves from occupation group 1-3 to stratum 15-17.
0cc group 6-9 is the one major, exception to this ranking primarily
because its SVP range.is so broad (SVP=3.5-8.0). If we delete this
19To simplify the analysis the evaluation was done only for
workers whose education was received outside of the south (i.e.School' south = 0).
Z.84
267
TABLE 6.7
Z/HC RATIOS BY OCCUPATION STRATA
Oc ipationStratum
White Males
Z
IntervalED HC
Intervala Ihtervalb Z/ED Z /HC
1-3 45.87% 13.84% 13.84% 3.31 3.31
5 23.82 21.83 28.61 1.09 .83
6-9 41.49 15.25 15.25 2.72 2.72
12-14 28.42 18.24 42.82 1.56 .66
15-17 12.40 42.62 114.86 .29 .11
All Strata 21.93 37.15 97.90 .59 .22
Black Males
1-3 75.80 2.84 55.32 3.64 1.37
5 74.85 15.01 26.06 4.99 2.87
6-9 49.23 16.08 35.21 3.06 1.40
12 -14 66.83 14.74 50.58 4.53 1.32
15-17 ha , na na na n/
All Strata 61.82 23.21 83.60 2.66 .74
aThe ED-interval is the earnings range expressed in percentageterms and estimated by evaluating each regression at mean values forevery variable with the exception of education (years of schoolcompleted) which is evaluated at ±lo around its mean.
bThe HC-interval is imated ad seriatum with all non-human
capital variables evaluated at their means and the human capitalfactors evaluated at ±-10for continuous variables and zero an,s1
for those that are dichotomous.
268
TABLE 6.7 (Continued)
White Females
Occupation Z ED HC .
Stratum Interval Interval Interval Z/ED Z/HC
1-3
5
6-9
12-14
15 -J7
All M.-rata
1-3
5
6-9
12 -14 na na C na / na na
15-17 na na na na na
41.54% 15.53% 15.53%1
2.67 2.67
54.63 11.58 11.58 - 4.72 4.72
35.44 - -- --
49.63 19.264e 39.11 2.58 1.27
28.22 ' 59.71 59.71 .47 .47
58.22 1Q.17 19.04 5.72 3.06
Black Females
72.47 23.24 52.60 3.12 1.38
69.78 30.35 51.44 2.30 1.36
98.31 16.71 "16.71 5.88 5.88
All Strata
1-3
5
6-9
12-14
15-17
All Strata
110.27 13 65
64.70 13.58
84.61 10.92
40.79 2. 35
36.47 44.57
59.56 32.08
63.07 30.33 64.10 2.08 .98
All Race-Sex Groups
42.53' 8.07 2.59
34.3U 4.76 1.89
10.92 7.74 7.74
53.24 1.83 .77
110.20 .82 .33
92.88 1.86 .64
M
269
special case, we find a monotonic increase in the size of the human
capital interval as we move from the lowest skilled occupations to the
professionals category.$ In occ stratum 1-3 the total HC-interval is
a mere 13.8 percent while it reaches almost 115 in the 15-17 group.
Roughly the opposite trend is seen in the earnings differentials
associated with industry and-stratification factors (Z-iterval).
The largest Z-interval is found in the, lowest skilled - category while
the smallest is found among the professionals.' Consequently there is
a combined trend toward smaller 2/HC ratios as one moves to higher'$
occupation strata. In the lower skill groups the largest differences
in earnings are associated with differences in industry and occupational
attachment while differences in human capital begin to play a
relatively much more important role only on the higher rungs of the _
,skill hierarchy. In the lowest skilled occupations the earnings
intrval due to non-human capital factors is more than three times1
as great as the range due to schooling, skill, and experience while
amongiprofessionals the size of the Z-interval is only 1/10 that
associated with human capital. Over all strata, those workers with
la more schooling, experience and SVP as well as geographical
mobility earn almost 'double (97.9%) the annual salary of workers
in similar industries who have la less education, experience, and OJT
than average and who have never migrated since childhood. Compared
to this range, differences in industry and stratification variables
generate an earnings interval only 1/5 as large. Thus clearly for
the white male workforce as a wholt, the primary factors determining
270
the distribution of earnings are related to human capital. '
Nonetheless, for those "entrapped" in the less skilled sector,
industry:aud'stratification factors are by far the more important
. variables. As long as the entrapment continues, increases'in human
capital will have little realized value.
Among back males the results are more ambiguous. There does
not appear to be any clear-cut trend in the size of the human capital
induced wage intervals over the range of occupations nor is there/a
trend in the Z/HCratios. While black men have relatively la ger
earnings differentials associated with industry and stratification
factors, their HC-intervals are correspondingly largerrleaving
relatively smaller/Z/HC ratios than white men in occupation strata
1-3 and 6-9. On the other hand no single stratum ratio is below
unity suggesting that even in the relatively skilled strata the
non-human capital factors play a substantial role in wage determination.
The Z/HC ratio of .74 across strata discloses that both human capital
and institutional factors are each of critical importance.
For white women the non-numan capital factors ,clearly dominate
the picture with a possible exception in the professional strata.
The human capital induced intervals are universally small in the
first three occupation groups and fact in the 6-9 stratum human
capital differences have absolutely no effect on wage differentials
at all. All explained variance in these earnings are a function of
industry and occupational attachment, the Z/HC ratio being mathe-
matically undefined. Finally for all strata combined, the industry
tr,
43-
and stratification factors measured ad conjunctum are more than
, three times more powerful than the human capital variables measured
ad seriatum. Thus we move to the very opposite of the continuum from
white men, suggesting that human capital differences are relatively
insignificant in determining the female personal distribution of
earRingswhile non-human capital factors dominate the field.
The results if or black women are similar to those of black men
with the exception of occupatisn group 6-9. The human capital
intervals are of generally the same magnitude as the Z-intervals-
in each of the individual occupation zroups and across all strata.
' AgainAt appears that both hinuan capital factors on the one hand and
industry and stratification factors on the other play important roles-,
in the wage determination process. Changes in either set of factors
Can be expected to hive a substantial impact' on estimated earnings.
In concluding we can turn to the results for the whole labor
force taken together. Here we find general trends which parallel,
i,
those for white men, but levels that are much closer to those found
for-each of the minority groups. With the exception of the non-
comparable 6-9 strata, there-is a monotonic downward trend in theI
1
z-interval accompanied by a -less regular upward trend in the impact _
\6of the human capital module. Together th y produce a concise
picture of the relative impact of the two sets of factors. Among
the least skilled workers in the economy, industry and stratification
factors producean earnings differential 2.6,times the size of the
human capital interval. This ratio falls (with the obvious exception
Z89
1
272
of occ group 6-9) until it reaches .33 among the highest skilled
occupations. Again this leads us to the conclusion that human
capital factors are of substantial import but primarily only in the
higher skilled strata. For the rest of the workforde, institutionalr ij
and stratificationfactors E.ie unambiguously important as independent,
and to a great extent primary, determinants of the persdhal earnings
distribution.
For the labor force.as a whole, taking into account the relative
population size in each strata, a comparison of the Z and HC
intervals in the "grand pooled" regression suggests that the estimated
impact of the industry and stratification modules is about two-thirds
the size of the effect of the human capital module. Both are important
° with .human capital having a-slight edge.20
Nonetheless the masSive-
-earnings differentials associated with (1) industry °and occupation .
crowding (2) differences in industry characteristics and
20It should be emphdsized that the range over which the human
capital factors are allowed to vary is by no means narrow. In the"grand pooled" regression, the 93 percent earnings differential isthe' total interval between two workers who have the following humancapital characteristics
+HC-interval -HC-interval
School =13.68 years(Junior College)
Institutional Training
Iligrant
40 years Experience
SVP = 6.73(2-4 years of OJT)
School = 7.86 years(Elementary School)
ft
No Institutional Training
Non-migrant
16 years Experience
. SVP = 2.97(30 days-3 months OJT)
273
A3) miscellaneous factors concerning imperfections in the functioning
of labor markets including pure discrimination, information barriers,
and lock-in effects are obviously too large to ignore. Contemporary---
labor markets do not appear to be particularly efficient in matching
workers with given endogehous productivity characteristics to jobs,.
requiring these talents. After controlling_ for human capital as best
we can, the evidence points overwhelmingly to the fundamental soundness-
of institutionalist and stratification hypotheses and provides
substantial evidence of the superiority of the personal earnings
distribution theory presented here.
VThe implications of these findings for manpower policy and
0
particularly the row -wage workforce are far-reaching. It is to this
matter that we next turn.
CHAPTER VII
A
CONCLUSIONS AND IMPLICATIONS
This study began with.ja relatively specific concern: to under -'
stand why millions of full-time workers earn so little that their
families become "working poor" in terms of the Bureau of Labor
.
Statistic'ebudget for "low Standard of living" or worse yet the
Social.Security AdministratfOn's poverty line.1
Even more,specili-
. . . Acally_ our concern was to determine
.
towhat_extent the low incomes ot-
-the working-poor are primarily the esult ofd inadequate human capital.4 ,
a
-- vs: the legacy of labor market imperfections.
Inevitably this relatively_narrow problem gave way.to much0
0 ,;
. , .
4,,
-_'b questions about the determinants of earnings for the labor
force-as a whole -and finally. prompted the Construction of a general
distribution theory, and the deyielopment of a.comprehensive data set
to test it. Ifhile,the results of our inquiry are, of course, not
11n 1967, the Social Security Administration's "poverty line"
-for a non-farm family of four was'$3,410 while the Bureau of LaborStatistic's "low standard of living" budget for. an urban family offour was $5,915. These figures can be found in U.S. Bureau of theyCensus, Current Population Reports: Consumer Income, "Characteristicsof the Low-Income Population 1970' (Washington, D.C.: Gov't PrintingOffice), Series P-60, No. 8], November 1971 Table M., p. 19 and U.S.Department of Labor, Office of Information, "Three Standards of Livingfor an Urban Family.of'Four Persons, Spring 1967" ("rashiogton, D.C.:Gov't Printing Office), March 1969.
tiz
274
- 292
a
275
absOltitely incontrovertible, the evidence from the regression analysis
seems more than sufficient to warrant some.important conclusions
about the American labor market and particularly about the market for
less skilled labor. There area number of policy implications in
the manpower area that folloW,from this analysis. .
46
.Jt would be highly' repetitive to recap all of the results
presented in Chapters V and VI, but we can reitetate the major con-.
clusions of those chapters and comment on some of the implications "
that follow from them. .Obviously with the space available we can-only
,-,outline some of these implications. A more in-depth analysis will have
to_ wait for another day.
By far the most important conclusion of our analysis is that the
American labor market is considerably inefficient in terms of matching
What we have called "endogenous productivities" to marginal prpducts
or wages. ,,Much of the, labor force appears to be paid at rates not
consonant with their measured human capital.? The result as"relative
underemployment" of large segments of the labor force, particularly
among minorities and less=skilled workers. Without altering an
individual's human capital it is often possible, at lest hypothetically,
to increase that worker's earnings significantly by only "relocating"
2We should stress the term "measured" once again, for-it isalmost certain that some forms of human capital have not been included
in this analysis which partly account some of the unexplainedvariance in-earnings., In addition we should note that individualpreferences have not been etxplicitly taken into account so that factorslike I'voluntary" immobility may also be responsible for some of the
apparent "inefficiencies" in the labor market..
"4:93
276
the worker from one 'industry or occupation 66 another. The wage
intervals we discovered for similarly qualified workers..are large
'enolph to make the difference between poverty and a so-called adequate
family income, roeexample, when -we hold; human capital constant for
occupation group 1-3, where many of the working po6r are found, we
find a wage range of $3037 to $6386, figures that biacket the poverty ,
yline and the BLS "low standard of living" budget. In other occupation
strata we find large 'human capital constant" wage intervals as well:
$2516 ins occupation grodp 5, $3202 in group 6 -9, $2350 in group 12-14,
and $2974 in the highest skill category. For the full-time workforce
-as a whole, the wage range due to differential industry-and
occupational attachment_ii $7614 vs. $4784. Thus a worker in thp laBor
-having "average" amounts of human capital but who gains access
to a "permissive economic environment" will earn almost 60 percent
more than a similarly qualified worker in a minority-crowded,
competitiVe, unorganized,- low profit-industry.
What is also clear from the. analysis is that different segments-
of the labor force face very different problems in the labor market-
In general, low :Incomes among white men athe result of inadequate
human capital, although imperfections in labor market information and
- , , .
possibly the "lock-in" effects of prohibitively expensive geographical
.relocation and non-vested seniority and pension rights appear to
promote significant wage differences among less skilled workers. Among
_white women, on the other hand, measured differences inn human capital
can explain practically none of the large wage differentials even among
4N-
Z94
277
relatively skilled strata. Our analysis indicate's, in fact, that 95
percent of the difference in earnings between white women and white
men is due to factors other than measured human capital. Much of the
total variance in our analysis is left unexplained, but that which
can be determined is disproportionately caused by imperfections in
the job market. The segmentation of the labo r market into "male"
and "female" job slots seems to play a crucial role in wage determina-A
Lion. For blaeA men and women, both the human capital and institutional
hypotheses are borne out in the wage determination process:
In theoretically specifying "imperfections" in the labor
market, emphasis was placed on the "crowding" hypothesis. As we=
expected, the evidence for crow-ling is substantial although not
definitive. To prove_crowding as a culprit in the wage determination
process, it Wouldhave been necessary to estimates -of-,.-: .
. .
the labor- supply and demand functions in=each industry and occupation._
Unfortunately, for all practical purposes, this is an impossible task.
The minority, employment variables we chose as proxies have the problem-.
of being substantially colinear with race and sex particularly because-r
sex-linked stratification is so pervasive. Never'heless the "crowding"
_
variables were often significant within individual race-sex groups
(including white men) after controlling for human capital and even
in a number of the cross race-,sex equations after dummies for race and
sex were added. Both of these tests suggest that crowding has an
independent effect on earnings.3
What is important to remember,
3Data from the Census Bureau's Consumer Income series provides
278
however, is that whether the STRAT mOdUle measures the specific forh
of discrimination known as "crowding" or some other form of discrimi-
nation is less important than the fact,that something to do with race
and sex is an extremely powerful determinant of personal income. ItO
is perhaps the,major reason for the lack of colinearity between
endogenous productivity and earnings.
'`Insofar'as there is evidence of 'crowding" it was possible to
divide its effect into "simple" and "complex" forms depending on
whether in addition to the stratification factors differences in
industry choracteristica had an impact on the distribution of earnings.
;The - compelling conclusion' seems to be that "complex" crowding_is the
general rule throughout but particularly so for the minority groups. I
Simple crowding explained wage differentials"for white men in occupa-// _
tiOn Strata 1=3 and 6-9 while-all black male, black female, and white
female groups (excluding-white women in occ' group 1-3) were typified
by significant industry as well as stratification variables.
Employment in a permissive economic environment of extensive oligopoly,4
high profits, and capital intensity added significantly to the earnings
corroborati'Ve evidence of "crowding" of white female labor. Since1955 'the ratio of full-time, full-year white female/white male wage andsalary income has secularly fallen as the labor force participationrate of white women has risen. In 1955 the ratio was .644; by 1968 ithad fallen td .586 and is continuing-to fall. In the absence ofcrowding--and assuring no divergence in human capital or intensificationof pure "sexist" attitudesithere is no reason to believe this ratiowould fall. The increase in female supply should affect white malewages as -well, if crowding is not operating. No other theory seems to
explain this phenomenon as well. See U.S. Census Bureau, Current .
Population Reports -- Consumer Income Series, P-60, No. 69, April 6, 1970,Table A-8, p. 86.
96
t
279
of minority group members, a finding in complete accord, with
Aiitibn 1-institutionalist theory.
0"e thing that becomes abundantly clear in the analysis, parti-
cularly in Chapter VI, is the dramatic change in the relative
importance of the human capital and non-human capital factors as one
moves from the low-skilled to the high-skilled occupation strata;
InduStry and stratification factors are universally dominant amo'g the
lower-skilled strata while human capital takes on a larger and larger
role as one proceeds up the occupational hierarchy. For the labor
force in occupation group 1-3 we found that the industry and stratifi-,
_cat-ion factors produce an earnings differential 2.6 times the size of
=
the-wage differential due to differences-in human capital. But among
the-highest skilled group tbe ratio falls to only .33 after a near
secular decline through the whole occupational range. At the top of
the hierarchy labor markets appear much more "efficient" in
allocating workers accordihg to their endogenous productivity
'characteristics.
There area number of more specific findings that bear repeating.
0-e of these is the statistical insignificance of institutional
training as a determinant of earnings for every group with the
exception of black males. For black men,
significant and relatively substantial in
12-14 in addition to the cross-occupation
the training variable was
occupation 'groups 1-3 and
regression. In no other
regressionwas this true. This may be due to the poor measurement of
this variable, or it might have some important content as we shall
144'*7' j
a.,
280.
later suggest.
On-the-job training, as measured by SVP, is an especially
powerful variable in the higher occupation groups and across all
-occupation strata, but there remains great confusiOn as to what this
finding actually proves. Because SVP is only obtained after access,
to a speclfic occupation is gained, leis difficult to treat it in
like manner to the other variables in the human capital module. If
occupational access is barred by discrimination or some other1
imperfection in the labor market, SVP may be better treated as a
stratification variable and its eff
1
ct counted here. On-the-job
training is therefore of critical portance in wage determination,
but it is difficult tosuggest h.Ow ocial policy might be develbped_to
deal win- it based on our analysis.
Finally we should note that we have found practically no evidence'
of "compensatory" wage payments for physically demanding, unpleasant,
or_dangerous work. While otir proxies for these factors are not
especially well- measured, we often find a negative rather than positives
sign on these variables. If we follow the signs on the coefficients, 0
, of.
we note that the-Fe area number of negative signs in the low-skill
, categories followed by inJignificant coefficients in the middle and
higher skill categories. The, one positive sign we find is among
relatively skilled white men. Only here is the compensatory theory
borne out by the evidence. In the lower occupation strata differences_
in working conditions may be completely overshadowed by the effect of
stratification while in the upper strata the true effect can be
Z58
281
0
measured because stratification plays a much' weaker role.
Alternatively it is possible that'differences in ability and skill
have not been held constant enough to pick up small, but nonetheless
existing, "compensatory" effects.
Theories to Explain These Results
The overall picture then is one of a highly:imperfect labor
market stratified by race mid sOc.. By no means is the human capital
theory disproved or completely rejected, but the general theory of
\-
personal earnings developed` here is clearly: superior in its ability. -.
to describe the parameters of the \earnings distribution.' Yet the, T .
"theory" is primarily only .a description even ailowing'for the4, r
analytic properties of the crowdin, g,\ hypothesis, The unansvieredio .
question is what dynamic is responsible for promoting such a labor
m doA ket structure and'then what cabe'done to alleviate its perverse
distributional and allocational effects., We cannot hope to give a
definitive answer to this gargantuan question, but we can attempt,some
brief conjecture.
We should note once again for the record that a strict human
capital theorist will probably deny much of the evidence presented
here and therefore possibly not see the need for an explanation at all.
Arguing that the human capital module is.misepecified and the data
inadequate is one possible way-to explair away=the results found in\
this analysis. There is no "measure'of innate talent and admittedly
there is an inadequate specification of interactions between human
;99
4
282
capital variables. But this, in our opinion,-cannot-ccouht for. 4
the apparent wage intervals we have-found associated With non-human
capital factors, particlarly after loading the analysis in favor of
the human capital explanation at every turn. We feel that our analysis4
clearly does ddmonstrate the existence of widespread imperfections1
which cannot, be explained away so simply. Assuming our resul_s}
generally carect we need to explain-them.
One possible explanation comes from radical stratification theory.
The large wage differentials we have found associatdd with race and
sex can be interpreted as consistent with the "diyide and conquer"
theory which is currently being developed.4
At considerable risk of
oversimplifying and thereby vulgarizing radical theory, the argument
can be-paraphrased. In order to keep the whole working class from
organizing en masse to overturn the capitalist order, the "ruling
class" his consciously devised institutions to prevent the developmenta
of subjective class consciousness among all workers. Racism and
sexism have been deliberately instigated to affect divisionsNwithin
the working class along these lines. In its "vulga treatment,
radical stratification theory looks to conscious-racist and sexist
hiring and promotion decisions by management as the major tools of the
"divide and conquer" strategy. More realistically, however, radical
4,For one version of the "divide and conquer" theory see David M.
Oqrdon, Richard C. Edwards, and Michael Reich, "Labor Markr7tSegmentation in American Capitalism," Conference on Labor MarketSegmentation, Harvard University, March 16-17, 1973 (mimeo); Also see,Stephan Marglin, "What-Do Bosses Do?" Review of Radical PoliticalEconomics, Vol. 6, No. 2, Summer 1974.
300
a
283
theory points to the roles of social and cultural institutions,
-particularly the schools and "bourgeois" family customs, in dividing
4the working class.'
Taken in, this broader context, radical stratification theory, ,
welbelieve, has much 'to offer in producing An understanding of the
overall income and wealth dir we experience in the United
States It is clear that massive differences in schooling and in sex .
roles are fostered in &Lir society which end up segmenting the labor
force into different occupakdon strata.5
Race; sex, and social class,
as we argued in the general stratification theory, can easily be seen
as the primary exogenous factors in determining the final distribution
of income.
But the problem in the present'analysis is much narrower in at
least one respect. Here we have held human capital constant and
asked they question how much of the variance in earnings can be
,explained by other factors. We therefore need a much more specific
theory which relates these human capital constant wage differences to
-factors that operate in specific labor markets, not necessarily the
social milieu more broadly defined. One obvious answer to explain
wage differences is pure 'discrimination on the part of firms. Another
(
5For an excellent treatment of this subject,'see SaMuel Bowles,
"Unequal Education and the Reproduction of the Social Division ofLabor," Review of Radical Political Economics, Vol. 3, No. 4, Fall-Wintet 1971 and Samuel -Bowles, "Understanding.lhequal Econoric
Opportunity: The Role of Schooling, and Family Economic Status,"American Economic Review, May 1973.
4).
301
284
is simply infopmat;on Imperfections which never get fully resolved.
But there is good reason to believe that such answers are
indeed too simplistic. An adequate theory.must do more than expltin
the existence of wage differentials which are not related to human
capital. Such a theory must also be able to explain why industry-
related wage intervals are largest in the lower-skilled strera while
at the top of the occupational hierarchy the huMan capital elements-
dominate. Extending the brief analysis in Chapter VI, an eclecticI (
theory can be-suggested which meets these requirements. leis based
on a combination of theories including (1) job competition (2) labor
market search (3) quasi-fixed factor and (4) statistical discriminatione
all of which are placed within a specific historical context. A
rigorous treatment_of this eclectic model can most likely,be
demonstrated, but for the present we must be content with simply laying
out the basic structure of the argument. One thing that is especially
significant about the eclectic theory is that while it is consistent
4
with the "crowding" hypothesis and radical stratification theory,1%
it does not rely on a "conspiracy" theory of capitalist institutions.
As in Chapter VI assume a job competition model where job/wage
slots arP.given exogenously, at :least in the short'run, and there are
fixed costs of hiring and training labor. The fixed costserise
with increasing job complexity so t there is a general positive
relationship between the degree of 'fixity" and occupation strata.
Alsb assume that information about potential employees is imperfect
;.,,,and involves procurement costs. Information on average group
0
285
characteristics, even if_impreciSe, is relatively inexpensive to
obtain while information about specific individuals is costly.
With these assumptions and the'additional one that firms
attempt to minimize total labor cost in an attempt' to maximize profit,
we can generate a theory fully consistent with most of our findings.
Firms will attempt to minimize the sum of direct payrhents to labor
plus search costs plus hiring and training costs: These costs are
not independent of each other for higher wage offers can reduce search
costs by increasing the supply of labor to the firm and greater search
effort can reduce training costs by providing a higher expected
probability of acquiring workers who can be quickly and efficiently
---trained. Wherever the training requirements for a specific job-are
-minimal we can expect that the rationdl firm will find it unnecessary
to invest heaiiily in search, for recruitment "mistakes" do not force
the firm to incur,lerge sunk costs. On the other hand, wherever
.training'requirements are substantial, the Cost of a recruitment error
is considerable. Therefore we can eXpect that there will be a
positive relationship between the degree of fixity in a particular
occupation stratum and search costs. To reduce the risk of large
.unprofitable sunk costs, firms will search intensively for recruits
destined for skilled positions while expending little search effort for
workers who are hired primarily to fill unskilled (low "fixity") slots.
In more specific terms, firms will investigate the individual
characteristics of their prospective skilled employees while using
rules of thumb or general search strategies to fill unskilled job
303
_
I
286
slots. In the latter case, many firms will resort to statistical
selection of one sort or another using supposedly objectively
perceived group characteristics in making hiring decisions about
specific individuals.6
While "rational" in the limited economic'
sense, such a strategy is ob'lliously prejudicial by definition.
One important question, pf course, is what group characteristics
are used for screening. Here is where the historical context
inevitably plays a'crucial tole.- The social and cultural institutions
--and belief syStems embedded in any society are marked by,substantial
inertia. 'Onqt firmly established for whatever reason, they tend to
be passively, if not actively, perpetuated. Without reviewing American
(or for that matter much of all Western European) history it seemsI
hardly necessary to "prove" that both blacks and-women have
hlstorically been relegated to disadvantaged positions in the labor,
force, blacks through involuntary servitude and racial segregation
and wren through family custom.7
Through the years custom and habit0
1.have produced some objective diffeiences in group characteristics as
well as (and, probably more importantly) induced lingering perceptions
of differenc4s which may have no basis in fact. Both 9f these no doubt
have a s bstantial impact-on recruitment patterns.
a \I
6See Kenneth Arrow, Some Models of Racial Discrimination,') op. cit.
7This analysis obviously begs the real question: why did he
racial and sexual institutions and beliefs develop in the first, place.Here "divide and conquer" theory suggests one possibility.
304
287
In the context of our culture, statistical screening then works4
itself out in terms of racist and sexist hiring procedures, not
necessarily out of an express desire to "divide and ednquer" or out of
a deep-seated commitment to white male domination of society (although,
both of these may be operating). Rather if firms have widespread
beliefs'about the expected probabilities of employee "success"--
whether these expectations be groUnded in fact or not--the result will
be stratification of the labor force.
t
To review, the lower the degree of fixity, the smaller the
potential cost of a recruitment mistake.Which in turn leads to
minimal search effort and a general tendency toward statistical
-disCrimination as the firm's search technique. The end result
inevitably is stratification of the labor force in the lower occupation
.47
strata. If a sufficient number of firms screen on the same character-.
istics, the result will be crowding and the development of large wage
differentials between groups in the economy. Once the initial
stratification has taken place, differential supply of on-the-job
training (SVP) may tend to exacerbate these differences. Also once
this system has been generated, it tends to be perpetuated. If the
screening procedufes seem to have "worked" in the past, they will tend
to become rules of thumb to follow in the future. Thus even without
pursuing a Conscious policy of "divide and conquer," a private cost-
minimizing system will tend to perpetuate non-human capital linked
stratification as long as labor market informati,on is imperfect and '
costly to secure. In the absence of government intervention, the
305
288
"social costs" of stratification will continue to be borne by
minority membersof the labor force. And these costs, asme have
amply shown, are often immense. Whatever private gain mii;hi come
from a "divide and conquer" policy if perpetrated may redound to the
"capitalists*" benefit even-without their active participation. In
effect, then,- a market system operating in an environment of
(1) ubstantial quasi-fixed costs for skilled labor (2) non-zero cost
information, and (3) a legacy of racist and sexist custom will tend to
produce a,"meritocracy" at the top of the occupational hierarchy and
racial and sexual stratification at the bottom.
What's To Be Done?
The labor market we_have uncovered is one involving-large scale
inefficiencies if one defines efficiency by a cOlinear mapping f
endogenous productivity characteristics and earnings. Yet we have
also posited a 'theory that the "inefficiencies" may be due to the labor
market operating the best it can given the context, of a supposedly free__"
market and limited information. If the market is to be moved toward
a more socially "efficient" and equitable allocation of labor, what
must be done?
Obviously a labor market which is segmented in such a complex
manner as we have discovered requires a multifaceted set of policies
to ensure equal opportunity in the labor force and reduce allocational
inefficiency. Wo single policy will be sufficient to redress the
stratification in the labor market. Without going into detail, we can
306
289
lay out a few areas in whith we feel policy must be directed.
One obvious finding is that although labor market stratification
is widespread, 'differences in human capital are still extremely
important. Within each race-sex group increased human capital in the
form of 'formal education, labor force experience, on-the-job training,
and migration all pay off interms of higher earnings. Yeethere are
great disparities that, remairi in the allodation of human capital \
between individuals particularly on the basis of race and class. This
8
QO
To redress the balance requires at a minimum equal. educational
has been amply demonstrated by other researchers.
opportunity if not compensatory educational programs for groups which
have historically been at a competitive disadvantage. In order to1
ensure equal labor market opportunity may in fact require unequal-4
educational opportunity, discriminating in favor of previously
discriminated against minorities. Quota systems and direct application
of affirmative action in college admissions, for example, are probably'
required. Other forms ofOluman capital may be equalized by providing
relocation allowances for those who can profit by moving from one area
8See for instance, Samuel Bowles, "Schc:ioling and Inequality fromGeneration to Generation," Journal of Political Economy, May 1972; W.Lee Hansen and Burton Weisbrod, "The Distribution of Costs and DirectBenefits of Public Higher Education: The Case f California, Journalof Human Resources, Spring 1969; Stanley H. Ma ters, "The Effect ofFamily Income on Children's Education: Some Findings on Inequality, ofOpportunity," Journal of Human Resources, Spring 1969; James S. Coleman,et al., Equality of Educational Opportunity (Washington: U.S. GovernmentPrinting Office; 1966); Patricia Cayo Sexton, Education and Income (NewYork: Viking Press, 1964); ancb Samuel Bowles, "Towards Equality ofEducational Opportunity?" Harvard Educational Review, Vol. 38, Winter
1968.
O
3 o7
P
290
to another and providing incentives for firms to ,give on-the-job
training to minority group'members.9
But the primary implication of bur analysis is that manipulation.
of non-human capital factors 16 also critical to addressing current
labor market Problems, particularly of the working poor. Insofar as
direct discriminatiOn is still widespread in the labor market, it is
clear that equal employment opportunity legislation must be extended
and forcefully implemented. As in education; affirmative action in
employment is an important tool in promoting social efficiency in
the labor market: More recent implementation of affirmative action
may have already begunyto excise the wage_differ,pntials betc.feen race-
sex groups. Clearly_such a direct approach to ending discrimination'
is warranted by the,results presented in our-analysis.
Beyond direct affirmative action, there seem-to be a number of -
roles the government can play in regard to statistical discriminatiori.
If,_as we suspect, screening is ,oftgn baSed on erroneous conjecture
"about group characteTistiCaplyegovernment can help to "correct the;
\-record." Such intervention in the market would not have the same
g
powerful effect f direct action, but it no doUbt should be in the-
ernment's policy tdol-box. Where substantial,discriMination is
ective," then the government must find the means for decreasing .
the rivate sector's cost of procuring individual job applicant
I
This latter policy suggestionshould be qualified foraccordingto the General Accounting Olfice, firms can take advantage of on-the-job training subsidies without providing much additional benefit todisadvantaged workers. See Chapter I, foptnote 8. . °,
3G8
291
information. This role could be played by a much more effective
public employment service which would have the funds and the
expertise to accurately screen individuals on the basis of relevant
job characteristics.10
The current Employment Service-has generally
(
failed in its attempt to bring low-skill workers and jobs together.11
More specific implications can be drawn from the analysis about
manpower training programs. We noted at the beginning of this inquiry
the: most social scientists and government officials have been
disappointed with the performance of manpower programs in the United
States. In the present analysis we find additional evidence that
institutional vocational programs have failed td have much of an
impact on earnings (althoUgh we-have.no information on their effgt--
on securing employment). The_ "training" variable is_significant_only__
for black males as far as the individual-race-sex equations are con-
cerned. Training is never significant for white men nor eithef group
of women. This result caused some consternation for we originally
suspected that if any group should benefit from institutional training
10See Richard Lester, Manpower Planning
(Princeton: Princeton University Press, 1966)Manpower and the Public Employment Service inYork State Department of Labor, 1966).
in a Free Societyand Alfred,L. Green,Europe (New York: New
110ne new picoe of evidence for -this conclusion comes from a
recent'study in the. Boston labor market conducted by the Social Welfare:Regional Research institute, Boston College. In virtually p11 of thefirms studied by SWRRI, the employment service was not considered areliable source for obtaining relatively less skilleliqabor and there-fore was rarely contacted about job vacancies.' Robert Hubbell-rand
Martha MacDonald, "A Study of Employee 'Recruitment _in Boston," SocialWelfare Regional Research Institute; Boston College, Workirig Paper,forthcoming.
309
vi
292
it would be white men. But here we find only black men apparently
gaining from these programs. One explanation, of course, is that the
result is purely spuriou, but there is an alternative that we prefer.
It has often been suggested thatomanpower training programs do
'little to increase the actual productivity of workers but play a
primary role,in the screening of recruits.12
White males do not gain
from this additional "screen" because they already are "screened into"'
the better occupations and industries by reason of their race and sex.
For blacks,_ however, training-plays the critical role of signaling to, 41
potential employers the special motivation that trainees may engender
. or appear to engender. In this case firms can use enrollment in e
training program as a way of screening in a few black recruits whhef
1.,-,-_-the normally operating racially-linked statistical discrimination ,_:-,.,..0
.1
cz: __- . ,
-sort-ens out all others. If this_is true, then institutional training
is obviouslyen important "human capital" variable for black men,
although its usefulness as a screening device might depreciate as thett
number of institutionally trained black workers increases. All of this
is but conj- ecture at this point, but it makes some sense within the
13context of the general stratification theory underlying our analysis,
12See Ivar Berg, Education and Jobs: The Great Training_Robbery.
(New York: Praeger, 1969).
13It has been suggested to me by several colleagues that *insti-
tutional training may even be a negative credential for white men ifemployers see enrollment -as an indication of labor market disadvantage-
mentA A "'good" worker should not,need a vocational training prOgram,might be the thinking of employers.
310
293
Finally we may conclude by mentioning the broadest implications
of-our analysis. What we believe we ultimately have shown is that
the distribution of earnings in the United States is substantially
arbitrary with respect to human capital. A large part of earnings
differentials have been shown to be related to non-human capital
factors so that the overall distribution of earnings can be described
as "unfair" with equal human inputs being rewarded with vastly unequal
returns. Much of the- justification for existing wage and income
differences ittributed to marginal productivity theory thus pales
before this analysis even if one has accepted the questionable premise
that a just distribution of income is one based on marginal products.
Imperfections are so extensive and their effect so deep that the
14
relationship between endogenous productivity and marginal product is
far from colinear.
'If individual policies of systematically countering imperfections
in the labor market cannot assure a solution to the distribution
problemwhich is very possibly the case--then it will probably be
required, at least in the short run; to resort to direct redistribution
- -of-income- via- negative income taxes or other forms of-income guarantee.
Such a redistributiOn would be far from Oerfect in-redreaSing "the
balance, but would be in general accord with the--policy implicationS
that flow from the crowding hypothesis. Under a negative income tax,
14See J.B. Clark for an early statement of the neoclassical
"just" wage doctrine. John Bates Clark, The Distribution of Wealth(New York: MacMillan, 1900), esp. Chapter 1 or 'Milton Friedman, PriceTheory: A Provisional Text (Chicago, Aldine, 1967), Ch. 10.
311
O
294
income would be.transferred from those who have higher earnings
in part due to segregation to those who have.been the victims of a
stratified labor market. In this case direct redistribution is a
surrogate for what would actually occur in the labor market if
barriers to mobility were reduced.
Thus the policy implications of our findings are extremety'far-,
reaching. They demand that policy-makers understand the need for-
wide7ranging intervention in the economy at the micro level in order
to move toward a more "efficient" and distributionally fair labor
market. Direct attacks on the structure of labor markets will often
_be much more effective particularly for lower-skilled workers than
attempts to remedy all problems through individualistic human capital
policy. All of this, of course, abstracts from even broader questions
of the control of the economy at large . . . but this is a question
to which I hope to devote my future work.
312
APPENDIX -A
THE DATA SOURCE
O
APPENDIX A
THE DATA SOURCE
The ability to test the general earnings model developed in
this thesis rests on the development of a comprehensive micro-macro
merged data set. This data set which is based on the 1967' Survey of
'Economic Opporthniq. was prepared over the course of one and a half
years by Professor Mary Stevenson, now of-the University of
Massachusetts-Bos'ton, and myself with the help of the staff at the
Institute-of Labor and Industrial Relations-Research Division at the
University of Michigan.
Three types of data were compiled, and merged to complete the
1 computer file, S1480,'which was used in all of the analysis. Data
from the-1967 SEO formed the basic sample. Information on personal
-characteristics flom this source wa,g suplacmented by-linked information
-from the panel portion of the 1966 SEO. The second type_of data
involved the compilation and linkage to. the SEO file of data on
=
occupational characteristics, the primary sources being the Dictionary
of Occupational Titles and the U.S. Census volume on "Occupational
Characteristics." The third type of data was macro information on
industry characteristics compiled from a broad range of sources. This
too was merged onto the SEO file so that the final 51480 tape contained
personal information on a 'large sample of workers with accompanying
296
314
297
0
data for each individual on that Worker's occupation and industry.
_
Thuis for every worker on the SEO file'we can ascertain such individual
characteristics As educational attainment and age, but also a-,goodO
deal about the training required.and the working conditions in that
worker's occupation and the profit rate' and the concentration ratio
--in that worker's industry.
The SEO Data Base
The Survey, of Economic Opportunity (SEO) for\1967 contains
information collected in a sample survey conducted in the Spring of
that year.1
The Bureau of Census conducted the survey for the Office
of Economic Opportunity to supplement the information regularly
collected in the-Current PopUlation Surveys (CPS) for February and
March of each year. The common items in the SEO and CPS include
personal characteristics (e.g. age, race, sex, family relationship,
marital status), last year's work experience and income. The CPS,
conducted every month, is the basis for monthly labor force statistics
including unemployment rates.
In addition to the normal information gathered in the CPS, the
SEO provides information on dimellsions of poverty not usually obtained
between the decennial census years. Of equal importance, the SEO
oversampled nonwhite census tracts to achieve better estimates of the
Ihis file is available from the Data and Computation Center,The University of Wisconsin, 4452' Social Science Building, Madison,Wisconsin 53706.
315
298
characteristics of the nonwhite poor. The national self-weighting
sample has approximately 18,000.households, drawn in the same way
as the CPS sample. The oversample in areas with,large nonwhite
populations has 12,000 additional households. The combined sample
yields over 57,000 adults, age 16 and over.
Working 40 weeks in the previous year and a minimum of 30 hours
in the survey week at one's "usual" job. qualified an individual as a
full-time full-year worker for the purpose of the present analysis.
Individuals who worked at least this much, but responded that their
present job was not their "usual" job were not included in the sample,-
for the hourly wage they report may have little relation to the
occupation and industry indicated in their survey record. All whites
and blacks in the self-weighting sample 25 years or older who
,qoalified as full-time full-y,c5i workers by these criteria were _
Automatically included in our analysis. Nonwhites who are not black
were excluded from the sample, since including Puerto Ricans and
Mexican-Americans in with the white or black samples would be
misleading because of different labor force characteristics.
Analyzing them separately, however, is impossible because of small
sample size.
In addition to the self-weighting sample, blacks in the supple-
mentary survey who met the full-time full-year qualifications were
included in the study population. The blacks in the supplementary
sample were found to be quite similar to the blacks in the self-
weighting sample according to an examination of means and Chi-square
316
299
tests. Thus, combining the. two samples did not significantly
.alter the qualitative results and the additional size added
immeasurably to the reliability of the estimates of black character-
isEics. However, the whites inithe "nonwhite" supplement were not
included in the analysis. Preliminary investigation indicated signi-
ficant differences between whites in the self-weighted survey and the
supplementary sample. Without the use of weighted counts, the addition
of the whites in the supplement would have qualitatively altered the
results for all whites in the population. Without the supplement,
the sample size for the white population is sufficient to yield
reliable statistics in even highly stratified analysis.
In all, there are over 57,000 individuals 16 years and older in
the 1967 SEO file. Of these, nearly 14,000 meet the full-time full-,
year and "usual" job qualifications. (See Table A.1) -Almost 10,000
are in the:self-weighting sample; the rest are blacks in the supple-
mental survey. Nearly 9,300 are males; 4,700 are females. In all,.0
there are nearly 9,000 whites and 5,000 full-time full-year blacks.
White males are the most numerous; black females are ldast numerous.
Nonetheless there are 1,867 full-time full-year black women workers
in this study. In all four race-sex groups it is then possible to
stratify the sample extensively and still maintain ample cell sips
and reliable estimates even after discarding individuals for whom
there is incomplete occupation and industry. data.
The 1967 SE0 yields information on all of the individually
measured variables needed in the analysis with the exception of a
313
300
TABLE A.1
THE SE0 SAMPLE POPULATION OFFULL-TIME FULL-YEAR WORKERS
Self-weightingSample n
NonwhiteSupplement Total
White Males 6257 6 6257Black Males 594 2242 3036
Total Males 6851 2442 9293
White Females 2736 0 2736Black Females 362 1505 1867
Total Females 3098 1505 , 4603
.
Total White 8993 0 8993Total Black 956 3947 4903
Grand Total 9949 3947 13896
measure of vocational training. A question about institutional
training was asked, however, in the 1966 SEO. Given that he same
addresses were interviewed in both years, a large percentage dl the
sample population wen- interviewed in both years. Given this fact
and the ability to link the 1966 and 1967 SE0 surveys, it was possible
to add information about vocational training to over three-fourths
of the 1967 sample. This was done.
The final variables used from the Survey of Economic Opportunity,
included:
Hourly Wage - The dependent variable is hourly wage (last week)
computed from answers to questions which indicate the number of hours
respondent worked last week and indicating the respondent's earnings
311
301
last week. Given the proviso of full-time employment at "usual"
job, this variable is somewhat cleaner than comparable measures of0 4
'earnings. The major problem is the possibility Of inflating the
straight-time earnings rate due to special overtime rates. To the
extent that overtime is "usual" in a given occupation, this would not
necessarily overstate the earnings measure we are after. The major
problem arises when in the week before the interview, an individual-
worker had an unusual number of employed hours. It was not possible
to take this factor into account. However it seems reasonable to
believe that it would not seriously bias the final results.
Schooling - This variable is taken directly from the question on
"highest grade completed.
School-South - This interaction term is computed from the answer
to the question on highest grade completed and the answer to the
-- question "region at age 16." A dummy value of 1 is assigned to the
region South.
-Trainia& - This variable from the 1966 SEO,was intended to
.account for institutional training of a variety of types. These
included: business college or technical training, alrenticeship
training, full-time company training, vocational trai n the armed
forces, other.formal vocational or technical training, of non-regular
schooling. A dummy value of 1 is assigned'if the.respondent had been
or was currently enrolled in any of these training programs. It was
not possible to ascertain whether the respondent had completed
training nor whether the training was related to subsequent occupational
.319
302
attachment.
Migration - This dummy variable was computed from the question
"age left last residence." A dummy value of 1 is assigned to a
respondent who was a "nonmover" indicating that this individual had
never moved more than 50 miles from his residence since age 16.
Experience - This variable was computed from questions on the
individual's age and highest grade completed. The actual variable
used was {AGE- SCHOOL -6) indicating the number of years out of school
and presumably available to participate in the labor force.
Union Member - This stratification variable was taken from the
question, "belong to a union." It was asked_only of private wage and'
salary workers. A union was construed as an organ_ted craft,
industrial, or professional union such as the bricklayers union, the
United Auto Workers, teachers union, etc. Fraternal or.social
organizations such as,Masons, Shriners, Elks, etc. were not considered
as unions. Similarly, professional organizations were not considered
as unions.
Race - This variable referred to "white" or "negro."
Sex - This variable referred to "male" or "female."
In addition to these factor's, the SEO provided occupation and`
industry codes for each individual which were used to merge occupation
and industry data onto each micro-record.
X320
303
Construction of Occupation Scores
Three kinds of occupation data were collected: (1) data on
education and training requirements (2) data on worPing conditions,
and (3) data on the1degree of minority employment. The first two
types were compiled with the use of the Dictionary of Occupational
Titles (DOT) while the third made use of the "Occupational
Characteristics" volume of the 1960 Population Census.2
The DOT is a vast store of data on individual occupations which
is rapidly coming into use by social scientists after being primarily
the domain of job placement officers. The Dictionary has information
on the general education and specific vocational preparation requirements
as well as data on working conditions for some 4,000 specific
occupation titles. The data for the Dictionary were collected and
developed according to job analysis techniques established by the
U.S. Employment Service. In most cases, the same job was analyzed in
two different establishments in one State and then in two different
establishments in another State. The findings of these studies were
correlated and job definitions prepared. As a result, information
presented in the Dictionary reflects the findings of-the U.S.
Employment Service from approximately 75,000 studies of individual
job situations.3
2U.S. Department of Labor,, Manpower Administration, A Supplement
to the Dictionary of Occupational Titles, 3rd. Ed. (Washington, D.C.Government Printing Office, 1966); and U.S. Census Bureau, UnitedStates Census of Population: 1960,20ccupational Characteristics,"Series PC(2) 7A, 1966.
ASuleiEpnenttothelarofOccuationales, op. cit.,p. vii.
321
304
Foftunately for economic research, the General Educational-, ,
Development AGED) and Specific ocational Preparation. (SVP) scales
for each occupation are based om functional or performance require-
inents. According to Sidney Fine,4
These are the requiremets determined by objective jobanalysis as necessary and sufficient to achieve averageperformance in the specific tasks of the jobs. Suchestimates try to focus on the tasks performed in relationto-the things, the data, or the people involved in those_tasks. ti
Thus for.each occupation an objective analysis of the job, not the
worker, is performed, ,yielding an independent measure of job content.
This is,'of course, of crucial importance for otherwise there would
be in the dataa tautoloditgi relationship between aoworker's
characteristics and the characteristics of the job he or she performed.
A "yeIrs of schooling'--t-Cale for GED was avoided when the
estimates were being developed. Instead the approach used was-to ask:
What basic-skills are pdople supposed to acquire as a result of(
general education, and what is the role of these,skills in jobs?
Three fundamental skills were delineated; reasoning,'mathematics,
and language. The requirements for each of these fundamental skills
were explored in a variety of jobs, and the scales currently in use
were developed to permit estimates from very low to very high require-.
ments. Again, attempts to translate these estimates into a time
4Sidney A. Fine, "The Use of the 'Dictionary of Occupational
Titled' as a Source of Estimates of Educational atil TrainingRequirements," Journal of Human Resources, Winter 1968, p. 365.
32249
t's
305
scale related to years of schooling completed in one educatidnal
system were resolutely avoided.5-
Each specific occupation was assigned a GED score on the basis
of these requirements. 'The rating system is reproduced in Table
reprinted from the Dictionary.
The SVP scale, unlike GED, is a time scale. It is described as
follows:6
Vocational Preparation: This is the amount of time requiredto learn the techniques, acquire informatiori, and developthe facility needed for average performance in a specificjob-worker situation. This training,may be acquired in aschool, work, or a vocational environment. It does notinclude orientation training required of even every fullyqualified worker to become accustomed to the special con-ditions of any new 4nb. Specific vocational training includestraining given in a-, of the following circumstances:
(a) Vocational education(b) Apprentice training(c) In-plant training(d) On-the-job training(e) Essential experience in other jobs
Each of these various types of training for an occupation was con-
verted'to a time Period and the periods were then adde' to provide
the,final estimate of required specific vocational preparation (SVP)
for each job. The SVP scale is reproduced in Table A.3.
In 'addition to the GED and SVP scales, the POT rates occupations
in terms of "Physical Demands" and "Working Conditions." The Physical
Demands scale includes data on required strength as well as the need
5Ibid., pp. 366-367.
6Ibid., p. 368.
323
306
TABLE A.2
'GENERAL EDUCATIONAL DEVELOPMENT
Level Rusoning Dote loptetnt Itathomotical Development Langugo Development-
6
4
$
1
Apply principles of logical orscientific thinking to a widerange of intellectual andpractical problems. 'Dealwith nonverbal symbolism(formulas, scientific equa-tions graphs, musical notes,etc.) In its most difficultphases. Deal with a varietyof abstract and concretevariables. Apprenend themoat abstruse classes ofconcepts.
Apply principles of logical orscientific tninking to defineproblems, collect data,establish facts, and drawvalid conclusions. Inter-pret an extentive variety oftechnical instructions, inbooks, manuals, and mathe-matical or diagrammatic+form. Deal with severalabstract and concretevariables.
Apply principles of rationalsystems' to solve practicalproblems and deal with avariety of concrete variablesIn situations where onlylimited standardization 'exists. Interpret a varietyof Instructions furnished inwritten, oral, diagrammatic,or schedule form.
Apply common sense under-standing to carry out in-stniction4 furnished inwritten, oral, or diagram.matic form. Deal withproblems involving severalconcrete variables In orfrom standardized situa-tions.
Apply common sense under-standing to carry'eut de-tailed but uninvolvedwritten or oral instructions.Deal with problems in-volving a few concretevariables In or fromstandardized situations.
Apply common sense under-standing to carry outsimple one- or two-stepinstructions. Deal withstandardized situationswith occasional or novariables In or from thesesituations encountered onthe job.
Apply knowledge ofadvanced mathe-matical and Binds-601 techniquessuch as differentialand integral cal-culus, factoranalysis, and
'probability deter-mination, or workwith a wide vari-ety of theoreticalmathematical con-cepts and make,original applica-tions of mathemat,Ical procedures,as in empiricaland differentialequations.
Perform ordinaryarithmetic, alge-braic, and geo-metric proceduresin standard, prac-tical applications.
Make arithmeticcalculations in-volving fractions,decimals, andpercentages.
Use arithmetic toadd, subtract,multiply, anddivide wholenumbers.
Perform simple ad.:dition and sub-traction, readingand copying offigures, or count-ing and recording.
Comprehension and expression ofa level to
Report, write, or edit artic:eefor such publications as'newa-papers, magazines, and technicalor scientific journals. iPrepareand draw up deeds, leases, wills,mortgages, and contracts.
Prepare and deliver lectures onpolities, economies, education,or science. /Interview, counsel, Or advisesuch people as students, clients,or patients, in sue matters aswelfare eligibility, Vocational
_ rehabilitation, me tal hygiene,or marital relatiods.Evaluate engineering teebnlcaldata to design buildings andbridges. ' t, -
Comprehension -vend: expression ofa level tO-
-Transcribe-dietation, makeappointmenti-for executive andhandle his personal mail, inter-view and screen people wishingto Speak to him, and write rou-tine correspondence on owninitiative. jr
:-Interviewfjob applicant/1A°determine work best suited fortheir abilities and experience,and contact employers to interestthem in services of agency.
Interpret technical manuals aswell a$/drawings and specifica-tions, such as layouts, blue-prints, and schematics.
Comprehension and expression oflevel to-
-File, post, and mail such mate-rial as forms, checks, reeelpta,and bills.Copy data from one record toanother, fill in report forms; andtype all work from rough draft
for corrected copy.Interview members of household
to obtain such information asage, occupation, and number ofchildren, to be used as data forsurveys or economic studies.Guide people on tours throughhistorical or public buildings, -describing such features as size,value, and points of interest
Comprehension and expression ofa level to-
-Learn job duties from oralinstructions or demonstration.
Write identifying Information,such as name and address ofcustomer, weight, number, ortype of product, on tags orslips.
Requeat orally, or in writing,such supplies as linen, soap, orwork materials.
Lioniplet ot "ptInciples of tat ionol systems' am Bookkeeping. Internal combustion engines-elect:1c wain, naval. boast!building, nursing, tart mannement. amp Wang
Source: A Supplement to the Dictionary of Occupational Titles,p. A-6.
324
-r
307
TABLE A.3
SPECIFIC ./OCATIONAL PREPARATION SCALE
9 - Over 10 years
8 - 4 to 10 years
7 - 2 to 4 years
6 - 1 to.2 years
5 -.6 months to 1 year
4 - 3 to 6 months
3 - 30 days to 3 months
2 - Short Demonstration - 30 days
1 - Short Demonstration only
Source: A Supplement to the Dictionary of Occupational Titles, p. A-7
to perform physical tasks such as ,climbing and balancing, stooping,
kneeling, talking, hearing, and seeing. We restricted our measure to
physical strength requirements using the DOT scale reported in
Table A.4.
TABLE A.4
THE PHYSICAL DEMANDS SCALE FOR STRENGTHStrength (Lift, Carry, Push, Pull)
Maximum Lift Frequent Lift/Carr
-1 - Sedentary (S)
2 - Light (L)
3 - Medium (M)
4 - Heavy (H)
5 - Very Heavy (V)
10 lbs.
20 lbs.
50 lbs.
100 lbs.
Over 100 lbs.
Up to 10 lbs.
Up to 25 lbs.
Up to 50 lbs.
Over 50 lbs.
Source: A Supplement to the Dictionary of Occupational Titles, p. A-7
325
308
Obviously this scale is insufficient as a complete measure of physical
demands on a given job. Any other measure, however, would, have taken
a prohibitively long period of time to compute.
Our measure of "negative" working conditions also leaves a good
deal to be desired, but again-it-was the only measure that could be
computed without resorting to a full investigation of each occupation
in detail. Each occupation can be rated in terms of whether the job
subjects the worker to certain working conditions including "extremes
of cold plus temperature changes," "extremes of heat," "wet and
humidity," "noise and vibration," "hazards," and "fumes, odors, toxic
Conditions, dust, or poor ventilation." For each occupation, we took
-the number of negative. working conditions as our measure for this
variable. This is surely a weak measure for it fails to account for
the intensity or degree of the negative conditions. A given job
which subjects a-worker to great hazard to life and limb may have only
one negative trait "hazards") while another job might sabjcct :
the worker to some humidity, noise and fumes and thus be counted as
having three negative traits But again we felt this measure was
better than no measure at all.
The problem with using the POT to obtain measures of GED, SVP,
physical demands, and working conditions is that the more than 4,000
specific job titles were not at the time of our research readily
transferable into the three hundred odd census occupation codes.
(Since completing our research the BLS has issued a Census Code -DOT-
conversion table..) It was thus necessary to develop our own conversion-
326_
309
routin- so that DOT characteristics could be assigned to each census
occup tion code. This was accomplished by matching,DOT titles and
six digit codes directly to the list of census occupation names, using
o r knowledge of the occupation st-ucture to make the match. This
rocess took literally months to complete and once done was checked
independently by a group of manpower experts associated with the
Institute of Labor and Industrial RelationsResearch Division at th-e
University of Michigan. Finally we were able to check our conversion
routine against a preliminary version of the BLS conversion table
and make final corrections. It is possible that even after this
painstaking task there are some errors in our conversion system, but
we have a high degree of Confidence in our results. One possible
problem arises in weighting the individual jobs in the broader
occupation codes. Again we had to use our knowledge of the relative
numbers of each specific job in an occupation so as to get an
estimated weighted average GED and SVP score for each census occupation.
In most cases this posed little problem for there was little variance
in GED and SVP scores across specific jobs within aQcensus occupation
code.
After each of the individual census occupations had been assigned
GED; SVP, 'physical demands, and working conditions scores, it was
necessary to group occupations'into individual "Occupation Levels."
This was done on the basis of narrow GED and SVP ranges. The result
was a 17 level breakdown which isreported in Table A.5.
310
TABLE A.5
"OCCUPATION LEVELS "1 BASED ON GED AND SVP SCORES
OccupationLevel GED Range SVP Ranp
1 1.5-2.4 1.8-2.4
2 1.5-2.4 2.5-3.4
3 1.5-2.4 3.5-4.8
4 2.5-3.4 2.0-2.4
5 2.5-3.4 2.5-3.4
6 2.5-3.4 3.5-4.4
7 2.5-3.4 4.5-5.4
8 2.5-3.4 5.5-6.4
9 2.5173.4' 6.5-8.0
10 3.5-4.4 3,0-4.4
11 3.5-4.4 4.5-5.4
12 3.5-4.4 5.5-6.4
13 3.5-4.4 . 6.5-7.4
14 3.5-4.4 7.5-8.0
15 4.5-5.4 6.0-7:4
16 4.5-5.4 7.5-8.0
17 5.5-6.0 7.0-8.2
Unfortunately even with nearly 14,000 cases in the SE0 sample,
this`stradfication by occupation level left individual cells often
too small for statistical purposes after further stratifying by race
and sex. Therefore it was necessary to reconstitute the 17 levels
into a number of "occupation strata;" This was the final step in
preparing the occupation data for linkage to the SEO file. Thp
final five occupation strata were grouped so that each individual
stratum had the same GED range (with the exception of group 15-17)
-and a broader SVP range. The proviso was that each Stratum have
3 28
311
sufficiently sized cells to allnw fo: extensive stratification and
yet yield statistically significant results. The five occupation
strata used in the analysis include levels 1-3, 5, 6-9, 12-14, and
15-17. OccupatiOn level 4 was e)o..litd9d because of singll cell size.
Occupation level 10 was excluded because it was totally dominated by
a very poorly measured specifi job, namely clerical and kindred
workers, nec. Occupation 11 was excluded from the final analysis for
the same reason, its dominant specific job Werag salesmen and sales
clerks, fec. We were so unsure of the CED and SVP scores to attach
to these broad occupation groups that we chose instead to leave rhem
out of the analysis.
The other information collected on census ,)ccupations included
median education, median age, percent black male, black'sfemale,
and white female employment. All of this came from the appropriate
1960 Census volume. It was added to the occupation file and appears
along with all of.the other occupation data in Table A.7. The list
of Census occupations organixei by occupation level with theirr\SE0
code numbs is found in
329
312
TABLE A.6
CENSUS OCCUPATIONS BY "OCCUPATION LEVEL"(SE0 OCCUPATION CODES)
Occupation Level 1
11974 laborers (nec)
07671 manufacturing graders and sorters
04324 messengers and office boys
07654 food, nut, and vegetable graders and packers
07693 packgrs and wrappers08803 laundresses, private household09813 attendants, recreation and amusement
09820 bootblacks09823 chambermaid, ad maids
09824 charwomen an cleaners
0983i elevator operators
09874 recreation and amusement'ushers
11960 carpenters' helpers
11964 gardeners and groundkeepers11973 warehousemen (nec)
10902 farm laborers, liege workers
09841 porters
Occupation Level 211972 truck drivers helpers
07674 laundry and dry cleaning operators
07653 metal filers, grinders, and polishers
07673 textiie'knitteis, loopers, and,toppers
07692 oilers and greasers (excluding auto)
07710 textile spinners,.
09830 counter and fountain workers09835 kitchen-workers (nec)
11971 teamsters
09890 service workes (nec)
Q7694 painter,s
11963 garage laborers, car washers and greasers07695' photographic process workers07632 auto service and parking attendants
Occupation Level 311970 lumbermen, raftsmen, wood, choppers
06535 upholsterers '
Occupation Level 404315 cashiers
08801 baby81,tteri, private household
330
313
TABLE A.6 (Continued)
Occupation, Level 5
07643 .ma ufacturer checkers, examiners09851 gu rds, watchmen, doorkeepers11965 lo gshoremen, stevedores04351 telegraph messengers07730 operatives and kindred workers (nec):06444 log and lumber scalers and graders09834 janitors and sextons09875 waiters and waitresses04304 transportation baggagemen04323 mail carriers04353 telephone operators64360 typists
05383 hucksters and peddlers07714 cab drivers and chauffeurs09815 bartenders04313 bill and account collectors07631 assemblers04320 file clerks
Occupation Level 607685 mine operatives and laborers (nec)09812 attendants, professional and personal service (nec)06494 construction and maintenance painters09810 attendants, hospital and other institutions09854 sheriffs and bailiffs04350 stock clerks and storekeepers06435 heat treaters and annealers07640 railroad brakemen07690 motormen--mine, factory, etc.07704 sawyers07705 manufacturing sewers and stitchers07712 stationary firemen07715 truck and tractor drivers09842 practical nurses04354 ticket, station, and express agents04343 shipping and receiving clerks
Occupation Level 7
07703 sailors and deckhands06415 cranemen, derrickmen, hoistmen09853 policemen and detectives07720 textile weavers
06425 road machinery operators04341 receptionists
331
;314
'TABLE A.6 (Continued)
04345 stenographers04352 telegraph operators06404 bookhinders06431 forgdmen and hammermen07641 bus drivers07645 bus and streetcar conducors07651 dressmakers and seamstresses07670 furnacemen, smeltermen4 pourers07691 motormen--street, subpay, el, RR'07713 RR switchmen09814 .barbers09843 hairdressers and Osmetologists /
Occupation Level 8:06513 metal rollers and roll hands07721 welders and flame-cutters06413 cement and concrete finishers06434' glaziers07630 asbestos and insulation workers07672 metal heaters07675 meat cutters09850 firemen, fire protection10901 farm foremen06512 printing pressmen and plate fitters01015 athletes06490 millers, grain, flour, feed, etc.
UCcupatiGn Level 907613 appren. building trades (nec)06401 bakers
06495 construction and maintenance painters
06501 paperhangers06505 plasterers06514 roofers and slaters06515 shoemakers and repairers
07652 dyers07602 -appren. bricklayers and masons06503 photoengravers and lithographers06411 carpenters
Occupation Level 1004312 cashiers
04361 clerical and kindred workers, nec.04302 library attendants and assistants04340 postal clerks
315
TABLE A.6 (Continued)
Occupation Level 1104325 office machine operators07650 deliverymen and routemen04305 banktellers04310 bookkeepers04333 payroll and timekeeping clerks06420 decorators and window dressers05396 salesmen and sales clerks (nec)
Occupation Level 1202212 farmers (owners and tenants)06492. metal molders01180 sports instructors and officials09825 cooks (except private hOusehold)01185 medical and dental technicians03252 RR conductors04314 vehicle dispatchers and starters04342 secretaries05380 advertising agents and salesmen05393 real estate agents and brokers06402 blacksmiths06410 cabinetmakers06452 inspectors04303 doctors and dentists att Adants
Occupation Level 1306545 craftsmen and kindred workers (nec)07634 blasters and powdermen06480 mechanics and repairmen (nec)06453 linemen and servicemen06523 structural metal workers06424. engravers01193 therapists and healers (nec)03254 floormen and floor managers, store06403 boilermakers01161 photographers02222 farm managers03250 buyers and department heads, store03265 ship officers, pilots, pursers03280 postmasters03285 purchasing agents and buyers (nec)06421 electricians
06454 locomotive engineers06460 locomotive fireman06461 loom fixers
333
316
TABLE A.6 (Continued)
06471 airplane mechanics and repairmen06472 automobile mechanics and repairmen06473 office machine mechanics and repairmen0647'4 radio and TV mechanics and repairmen06475 RR and car shop mechanics and repairmen -06491 millwrights.06510 plumbers and pipe fitters06520 stationary engineers
06521 stone cutters and stone carvers06524 tailors and tailoresses06525 tinsmiths and coppersmiths06530 toolmakers, die makers setters
07604 apprentice electricians
07605 apprentice toolmakers and machinists07614, apprentice metalworking trades (nec)07615 apprentice printing trades07701 power station operators06430 foreMen06465 machinists06502 pattern and model makers06405 brickmason and stonemasons
Occupational Level 1406423 electrotypers and stereotypers06451 jewelers and watchmakers06470 mechanics and repairmen07610 apprentice mechanics, except auto06414 compositors and-typesetters
Occupational Level 1503260 publiC administration inspectors01074 draftsmen01104 funeral directors and embalmers01164 radio operators01190 electrical and electronic technician01191 other technicians01192 technicians (nec)
04301 agents (nec)01072 designers01154 personnel and labor relations workers01182 teachers--elementary schools
05395 stock and bond salesmen07642 surveying chainmen, rodmen, axmen01102 farm and home management advisors01111 librarians
334
317
TABLE A.6 (Continued)
01150 professional nurses01160 pharmacists01163 public relations men01174 statiticians and actuaries01181 surveyors01183 teachers--secondary schools01184 teachers (nec)
03251 buyers and shippers, farm products
03262 building managers and supeiintendents01090 metallurgical engineers01084 ,industrial engineers01093 engineers (nec)'01012' airplane pilots and navigators05385 insurance agents and brokers
Occupation Level 16
01014 artists and art teachers01073 dietician and nutritionists01171 social and welfare workers01001 accountants and auditors01120 musicians and music teachers03253 credit men03275 qficial--Lodge, Society, union04321 insurance investigators01085 mechanical engineers01082 /civil engineers01170 /religious workers010911 mining engineers01080/ aeronautical engineers01075) editors and reporters01085 electrical engineers01165. recreation and group workers
Occupation Level 1701020 authors
03270 public administration officials--administrators01173 psychologists
01030 college presidents and deans
01103 foresters,and-conservationists01013 architects01021 chemists01023 clergymen
01031 faculty--agricultural sciences
01032 faculty--biological sciences
335
318
TABLE A.6 (Continued)
01034 faculty--chemistry01035 faculty--pconomics01040 faculty--engineering01041 faculty--geology and geophysics01042' faculty--mathematics01043 faculty--medical sciencesr1045 faculty--physics.01050 faculty--psychology01053 faculty--social sciences,.(nec)01054 faculty -- non - scientific subjects
01060 .faculty--subject not specified01071 dentists01081...chemical engineers01105 lawyers and judgesb1130 agricultural scientists401131 biological scientists01134 geologists and geophysicists01135 mathematicians01140 physicists01145 miss. natural scientists01162 physicians and surgeons01194 veterinarians01172 economists
3C6
._7
319
The columns in Table A.7 refer to the following:
Column Datum
1 SEO Cens-Js Occupation Code ,
2 Physical Demands
3 Number of Negative Working Condition Traits
4 GED SCORE
5. SVP SCORE
6 Median Years of Schooling - -Males
7 Median Years of Schooling -- Females
8 Median Age--Males ,)
9 Median Age -- Females
10 '. Imputed "Occupation Level"
11 Percent Black Male Employment
12 Percent White Female Employment
13 Percent Black Female Employment
3,
/
...
SCIri _JO
9'0 601.'0 0'0 'LI. 5Er 10E S91 ZLL 04 09 0 PC,0C 5E0'0 010'0 'Ll S'EE 7'hf 591 7-L1 0.9 0'9 '0 720.0 11770 0E0'0 'L1. S'6Z E E 1791 LL o8 0.9 0 0'0 0E0'0 100'0 'Ll S'Ef 5SE 5'91 91 3"11 ic/ '0 910'G 057'0 SZO0 'Ll LZI: FnE 5291 9'91 OR 0'9- '0 0'0 190'0 OZ00 'Ll S'ff 6'Pf 5'91 9*9I 34 0 070'0 7950 PZO '0 *91 Z '9E Pbi f 0.0 O'S '0 61'0'0 018'0 900'0 'St 01E 7'91 L'91 C'S '0 00'0 1E00 100"0 'Ll L SP fct 0%1 PL 0.0 0'9 '0 013'0 150'0 1.00*0 '51 9'£S P'Zh L71 "r1 0'L S't) 7 1.00*(1 LZ00 P00'0 *Li 4'61 '71 6'L 3'9 '0 stip° Ent*o 6E0'0 '51 6'9f 7'0E Sql n'at O'L O'S 0
........ I. *** 0'0 600'0 L00'0 'St CAE l'ff ''II (al 7'1 7'S 'C
o'n zoo.° r00'0 'at L'6E 6'9E E'r1 5'00 OP 7'S '7
0'0 SI0'0 LOO*0 'Si L.60 0.0( 6E0 c.qi ht. U'S '0 9'O £00'0 100'0 '91 L'6E A9E 6'Et 1'11 R'L I'S .'0 0'0 1'70'0 £00'0 'St col 9'Zt rst f'L t'S '3 0'0 600'0 500'0 '91 9'hr E'9E 7'S1 7'91 P'L h'S '0 0'0 h00'0 80C'O '91 L6E EOh 6'E0 1'01 0'9 I'S '0
P*0 1110'0 100'0 *L1 L'6E 0.0E 6'EL /91 C'R 0'9 '0 O'C' StO'C 900'0 '01 L'6F 6'SE 6"'EL (-01 L'L r'S '0 C09'0 E91''0 S00'0 '91 S'6E 17'6C 9ht 0'91 S'L JOS '0 100'0 LS0'0 010'0 '151 L'ec 6'71 6'71 O'L S'h '0 971'0 6PL*0 LI0*0 '01 l'hh 69£ 7'Et L'71 S'L i'n '0 cod'o PLI0 900'0 *S1 S'th ut% n'ct c'Et t't 'o
1.01'0 070'0 920'0 'LL 5' #01 6',07 0'91 rLt 0'11 0'9 '0 910'C FIGL '0 800'0 'SI 6'87 t'67 S71. OL S' '0 LoO 6I2*0 110'0 L'1,/ A'PE P'Ll PL 0P 0'9 '0 rice 767'0 910'0 'L1 9*th E07 fr'Ll 11-LL 0'9 0'9 '0
0 zn0 6L70 070'0 'L1. 991 P'017 h'LL PLl 0'8 0'9 '0 00 M*0 0'0 'LL P'LL tLt 0'0 0'0 'C 0'O 780'0 0'0 'Ll a. n'Ll t'41 0'0 0'9 '0 "600'0 hOZ'O 010'0 'Ll O'Lf h'Lt h'L.1 O'R 0'9 '0
r1'0'0 9E0'0 E10'0 'Ll "'11 '09 P'a '0 L00.0 191'0 070'0 'LC l'Oh S'Oh h'Ll h'Lt 0'9 3'a '0
E10'0 int.° St0'0 'Ll L'rh 10tE h'Lt t'LI 0'9 C'0 '0 0'0 .970'0 Et0'0 'LI E'9£ h'LI h'Ll 0'9 9'9 '0
700'O OZO*0 ht0'0 'LI ** '"QE h'Ll 0'0 0'1 '0' 900'0 190'0 710'0 'LI **
' b-nt, hLI 1,11 0'9 3'9 -e cloco two 'Av.° -Lt *** hLt btt n'n O.4 '0 90/-0 551'0 'ITO '0 'LI 6'"f Z t'Ll hLi 3'9 ''9 '0 O'C 900'0 8E0'0 'Ll 07 P'LL bLL C'9 '0 c000 917'0 6E0'0 hf,h LLn nL hLt 7'9 9s '0 coo 610'0 L90'0 LL Vim 7.11 P71 I'LL nR 0-11 '0 0'0 160'0 4 (00 '0 'St E'rIS r'Sh 7'S1 511 3'L 0'S '0
E 9L00 6 1.0-0 L1. LEE O'LE 5'91 0'9l 0'11 0'9 '0 P1000 GLZ0 L 0'0 'LL Cm, 6PE 0'91 9251. ;L S'S 0 5000 SLO0 PS 0 '0 '147 5'71 5'S E'E '0
900'0 I1E'0 11 '0 '91 E'Rr 0tr l'St CFI S'L S'h '0
0'0 120.0 sno 6 -LI e.0'°1 O'S 1.00C 500'0 £00' *S1 ** .'9r * S'rt 3'L 0'S '7 h1VP 917f'0 E10'0 'St l're "RE t'El l'hl O'L O'S '0 E00'C Z91'0 SO0'0 '11 9'7h 0Ch R'71 r'r1 0'9 0'S 'n
11,0,4 11. is -
t
'1
7 'I '2
t/ '7
'2
'Z
'2
'1
'7 '
'7 'I,
'7
'1
'1
'2
'1
i '7 '2 '2 7 '2
'7 'Z
'7
'7
'7
'Z
'Z
-z
z 7 '7 '1
f -
'1
'E '7
't
'7
'7 '1
c^11, 011 cill "CIL 1E11
OF 1.
071 lilt S011 hOlt E011 2011
01. E601 7601 L601 0601 .
SP31 1000 E1131. ZPOL 1P0( 0001 SLOI tLOL fL31 7L31 1101 0101 1901 hS01 ES01 2501 1501 OSOl ShOt EhOt ZhOt 0101 OhOl. SrOt qrot. 7E00 1E01, OfOti (7011 77014 1201, orolt slot'
11011 71011 01011
10011, S6601 0000£
(ET) (zT) (TT) Cot (6) (8) (t) (9) (c) ('7) (E) (z) (I)
aaop oas a ma NoLwan000
auvI
OZE
feS*0 900'0 '11 uer rah 7f0'0 701.'O 110'0 It 9'Of 1'97 h10'0 AS1'0 571'0 t'61 Cf7 700'0 610'0 101'0 'S 7'fb C'Ob rove 1710 b00'0 '91 C'Sf e'bf SSO*0 (6L'O 910'0 'S 0'67 h'LZ 900'0 070'0 1700'0 'h J'/1/
110'0 001'0 170'0 *71 7'011 S'Ob oto.o 081'0 610'0 '5 0'7h R'LE 1170'0 RSV() L00*0 '01 1.'9C t'97 LOCO 070'0 700'0 'ft 0'6C 6'PE 700'0 909'0 f00'0 '11 L'7t 9'0C /100'0 610'0 701'0 'S S'7b 670'0 tC6'0 £00'0 '71 L'7( 7'lf 5'410'0 LOCO '610'0- '01 l'lf 1'77 500'0 171.1'0 110'0 '51 h5W urn oe.s. * 700'0 760'0 SO0'0 'ft R.Th. 7.7h 100'0 70r0 100'0 'Et O'TS 5'05 100'0 660'0 1'70'0 '91 A'91b L'Lh 900'0 Sel'O 010'0 'Ll 9'Lb f'Ph 100'0 (00'0 1.00.0 11 L'Sh 010'0 19C0 100'0 'St 0'9S L'hy S00'0 7110'0 PlO*0 'St ALII h'sh 170'0 7.Lh'0 P10'0 'ft L'7.b O'C 6(1'0 b00'0 '91 6'7h O'AE 0'0 f00'0 S00'0 '71 rtS 100'0 L10'0 000'0 'St 'Rh 700'0 Lt7'0 b00'0 'ft C'S4 6'011
700'0 1E0'0 hZ0'0 'Cl S'fh 900'0 Oh0'0 (90'0 '71 L'tS 7'Fb
0'0 1110'0 L10'0 'Li 7'6( h700 505*0 1E0'0 'Et L'14( f'7h 510'0 077'0 Of0'0 '51 6'51 6'7C 1100'0 771'0 1110'0 'St E'9( L'7C f00'0 9b0'0 616'0 'St 6'9F /'Of 7h0'4 SLYO 1E0'0 '71 flf, Vhf LF0'0 ofvo
1LS*0 hrvo
010'0 F70'0
'51
'St
Vrt, 11'14
b'LE 6'4C
LLO*0 LLCO ht0'0 '51 S'bh h'hE 100'0 AC0'0 100'0 'St C'L( 41f 910'0 00(0 Sf0'0 '71 A'tf O'ft 500'0 197'0 ffO'O 'Ll 9*Ch 7'011
LOO*0 b0E0 b10'0 'St 9'0h ('RC 600'0 00i'0 710'0 'Ll 6'61 h'cir
TO0'0 1E1'0 010*0 'Ll L'6( 9'6f Fr9'0 145'0 100'0 '91 Chn C.'PE
S10'0 665'0 710'0 '91 S'Sh CLF Ab0*0 SLUO h11O*0 '91 fl'ff
1100'0 A60'0 110'0 'St rot f'5E b00'0 ((7.0 900'0 '51 L'9C fqh 701'0 190'0 070'0 'Ll 9'1h l'fb Ervo 911'0 610°0 'ft q'Ob Z'hr 700'0 R[0'0 L10'0 '51 1'6C b'Sh 400'0 ("67'0 1400'0 "St 6'lb 7'lb 500.0 -(01'0 500*0 Aph E03'0 Oh0'0 100'0 'St 4'1h
i b50'0 1716'0 [00'0 '51 l'6( 50h
62C
17.71 (-.?t 6's 0.11 *0 .1 fret's'
h'71 S'h S'C '0 '7 czfnr .A11 901 0'7 3'7 '0 '7 h/f0C 7'71 f'71 irr *0 't mho 9'71 4'41 O'R l'S '0 1 17111(
Z*71 h'71 O'f O'C '0 '7 07f110
+7'71 hE '0 'h S1f110
0* l irn 0'7 irtt .0 .1 .httho 1Z1 f'71 r.f .0 '7 11010 0'7 3'71 O'f S'C *0 1 ZlfbO Sfl L'71 O'S O'b '0 1 01fb0 '5'71 L'71 O'S O'h '0 '7 SO(b0
s'Ol O'C Q. '0 'C hOrb0 S71 S'71 0'9 7'h 1 *7 10fh0 lfl O'h O'h '0 'f 701110
s'n 1'71 O'L 5'h '0 '7 lOCh0 067C0
ci'n 71 O'L 0'4 '0 't 5117E0 E'71 S*71 O'L O'h '0 1 047f0 S'M 1'71 O'R O'S '0 't SLZFO L*71 6'71 S'L 5'S '0 1 °UFO
S'Ot O'L O'h 1 '7 So/CO ('II h'6 O'L O'S '0 'Z. r97f0 h'71 t'71 5'4 S'h 097C0 S'll 0'71 S'I O'h '0 '1 hS7C0 S71 rrt 0'f1 VS .0 1 r57r0
5'01 0'9 O'h 1 T. 717(0 0'11 O'L O'S '0 '7 1S7f0
h'71 C'L O'h '0 q OSZCO O't' O'h '0 '7 77770
"0 CA S'S S'f '0 71770 S6110
'Ll O'A 't '7 66110 1'91 h'911 S'9 O'h '0 '7 F611C /71 0'71 0*1. S'h 'C '7 76110 9'71 P*71 O'L 5'h '0 '7 16110 C'71 1'71 C'L S'b '0 '7 06110 7Et 0'71 0'9 O'h 0 '7 S911( 1'91 h'91 3'1. O'S '0 '7 hell( 9'91 7'Ll O'L 0'5 '0 '7 fAll( bqj o'Ll C*9 O'S '0 '7 7.011(
cri O'L Vs *0 '7 10111
S'91 5'91 0'9 SF '0 'E 0811( 5'91 O'Ll 0"9 '0 *1 Silt( O'ft 1'91 3'L O'S '0 1 ht.C11
h'il S'il P't P'S '0 't tLtIl 6'51 A'91 O'L 0'9 '0 't 71,111 q'qt coqt S'L 0'5 '0 1 11.111 toft 9'L F'S '0 ntAtt L'hl 1'51 S'L b'S '0 1 59111 UZI 9'71 O'L S'h '0 '1 11911.
b'St 1'91 O'L O'S '0 1 09111 b'Ll 5*/1 'O'R 0'9 '6 '7 7911- b't 4o71 O'L 0'4 '0 '7 1911" 7'41 7'01 O'L O'S '0 .7 09111 t'fl 6'51 S'9 0'5 'n 't bqtt
r'Lt 3'9 3'9 '0 'f ES11. C'Ll O'L O'S '0 1
* 1511 ZEt Cft O'L 3'5 '0 '7 OSIl
(cT) (zT) (TT) (9T) (6)' (8) (L) (9) (S) 07) (0 (z) (T)
(panuT4uo0) L 'V TIUVI
TZE
ZOO*0 SZO*0 777'0 Z1 0'0 100'0 610'0 'EL
0'0 SO0'0 S90'0 'A ZOO*0 510*0 150'0 o 'FL 0'0 100'0 Ch0'0 '11
700'0 St0'0 7h0'0 'Et
100'0 U0'0 710'0 -rt ioire rovy 940'0 'EL
100'0 1110'0 L70'0 'EL 100'0 100'0 170'0 'hi 100'0 C10.0 SZO*0 'EL LoVo novo zovo -11
040 C00'0 rZ0*0 'CI 0'0 700'0 Z00'0 'FL
tO0'0 610'0 110'0 'CI
t00'0 POO*0 tIO*0 '7L
600'0 090'0 010'0 'hi
600'0 1100'0 CP0'0 'S
200*0 L10*0 690'0 '9
n'D 770'0 710'0 'R
0:0 two L70'0 'CL
r00'0 8E0'0 190'0 'L
700'P 990'0 410'0 'EL
0'0 h00'0 LC0'0 .L
600'0 h111'0 120*0 'ft
0'0 L00*0 000'0 'hi 0'0 L00*0 010'0 'CI RZ0'0 P7h0 rZO'n 'IL
LC0'0 000'0 L60'0 't
hoo'n r10'0 07C*0 'ht 0'0 700'0 6LZ0 '8 0'0 1O0'0 ShO*0 '6
100'0 710'0 970'0 '7I
tovo row° IWO '1%
own 715'0 670'0 L l'i'0 100'0 .9Z0'0 'CI 0'0 900'0 670'0 '71
117.0'0 0E1'0 790'0 '6
600'0 R6E*0 600'0 'LL *0'0 090'0 S00'0 'St 900'0 P77'0 1'10'0 '71
100'0 ZhO*0 th0'0 'h 900'0 560'.0 1110'0 'St 070'0 P4S.0 920'0 'S L60'0 616'0 ZOO*0 :S
0"-0 CZO*0 0'0 'R
100'0 O51*0 h00'0 'ZL
ZZO*0 595'0. (t0'0 '01
SSO*0 SaR*0 h00'0 'S
100'0 617'0 900'0 '9
C70'0 IC6'0 Z00'0 'S
100'0 SLZ*0 SO0'0 'I_
1110'0 090'0 L91'0 'S Z70'0 471'0 1P0'0 '9 610'0 L76'0 700'0 t 00*0 9L0'0 001'0 'q htn'n 1S6'0 100'0 '7L
S70'0 Sh6°0 700'0 't ewe IWO hht'O 'OL
ckc
CRC n'th r'01 n'n S'S S'C s *r 7c.h901
0'Sh 6'6 O'L 0'4 1 " 16h90 6'Eh .6.P I'9 1E '7 'Z
01 0',01 -7'7h 9'01 hOt 8'9 LE '0 '1'
CIOP:70
S'6C E'70 VAL P'Ot t'L O'h '0 '1'
r4Oh I'll S'L O'h 'C .0
111 O'L0 l'h O'L O'h 7 "tt
7'04 ,'4f. 6'tt I'7t O'L On :0 '1'
_to7t O'L O'h -'0 '7 vrr hClf 901 f6 3'L O'h 'Z 'r
6'111 L'll 9'11 7'71 O'L Oh 't -r
yfh 9 O'L 0'0 'I 't 1qh90 7fC o L'LL O'L n'h 't 'Z
9'6C hhh S'll r'LL S'L O'h '0 'L
6-LE. /*Eh R'Ot 66. L'C L'Z 't 'C
S'4E i-rr t'7.1 1-7t S'9 CPC % 't
h'Lh 6'6 0'9 O'h '0 '1'
ihhtr°0
0Shq0
::: i'tS 4.44...11'6 O'L O'h 'L '7
t'bh 11'6 O'h O'C 'C 'C Srh90 11E100 L'SC .9'11t 0'9 vr -(1 *C
605 0'01 O'L n'h '0 'L 7rh90 c'f h f'11 O'S O'r 'Z. '11 tr490
9'hh L'Ilh 6'01 S'LL rt. O'h 'Z 'Z Ei L'Oh P6C O'll 0'71 O'L R'C '0 '7
6'6E t'f, 6'0 O'E 't 'E
nth U O'P S'C 'Z '1'
S'Lh O'Lh O'Lt R'11 O'L O'h 't '1
S'th S'RC h'Zt fIt O'S O'h '0 :r g * 4b'th ..6n S'h O'C 't 'I
S'Oh r on 7:71 t'71 O'R O'h '0 'Z
47'0h 9'11 3'9 O'E '7 'I Cif L'qr L'rh 0U1 66 0'1 O'f 'Z 'f
tlhlg
O'fh 0'01 3'9 O'h '0 '1'
4 O'PC W6 C'L YE' 'Z 'C
:::::
O'Ch L'6C s-tt rzt vs 7'0 '0 '7
S'Sh UZI, L'4 Z6 e'L 'O'E 'l 'f t'LS S'P D'q 3'h 'Z -t,
i°!;(9!
ionno
hqh h'RE 6'6 POI 3'S O'r '0 '7 I'Sh .4e.,..9.6 S'9 O'h 'Z 'h
LLh Lhh 8'7L S'hl S'9 O'S '0 '7 snrso L'811 h'OS 9'71 L'Z1 0'9 0'0 '0 'Z r6CSO L'LZ L'St 0'01 0'6 O'Z O'C '0 '7 06CS0 q'Ch or, 9'71 1.71 0'9 L'h '0 't SACS° E14 6'Sh l'T,t Z'OL 0:E O'C '0 'C CPESO 8'8C C'LC t.'It 0'71 O'C O'C '0 '7
O'Lh smt9'11 0'9 1 nr '0 'Z = L'6C 0'61 8'71 r'Et 0'9 O'h '0 'Z oorso Z'RE P'11C t'ZI hZt 7'4 O'C '0 'L
L'SE O'LC 7.'71 E71 or O'C '0 't
R*67 O'Oh CIL ('It O'h I'C '0 "Z
9'hh O'Oh UZI 7*Z1 O'S n'r '0 'L
R'01 h'Ot Of S'Z '0 'Z
1(1111!!!!
ISOM
0'6Z 9'RZ S'Zt S'7t O'r O'C O 't
E'6E 9'hC 0'71 11'11 O'h vr '0 'L
ZSCh0
6.7r CRC 9'7t o'71 O's 0'f *0 el
OSCh0 ShrhO
n'fiC L'IC 5'11 Aet S'E O'C '0 'I'
Z'Sh P'Oh h'tt 6'71 0.11 0'4 '0 '7
C7101M nhr vrn 9'71 0'71 vn O'h '0 't
Crr SW 5'71 c'rt OS Of '0 't
(ST) (n) (TT) (OT) (6) (8) (L) (9) (g) (h) (£) (Z)
(PanuT4u0b) L'v
ZZC1
CO
323
/TABLE A.7 (Continued
(1) (2) (3) (4) ,(5) (6) (7) (8) (9) (10) (11) (12) (13)
06'493 2. 2. 4.0 6.0 10.5 47.1 12. 0.030 0.022 0.001
04494 1. 0. 2.8 3.4 12.0 11.4 39.9 40.3 6. 0.014 0.145 0.007
06495 2. 1. S.0 7.J 9.1 11.7 45.5 41.9 4. 0.068 0.018 0.002
J4501 ' J. 1. 3.0 7.0 0.6 d.a 51.3 47.9 9. 0.108 0.126 0.002
0659 J. 1. 4.0 7.4 12.0 r 42.2 13. 0.007 0.016 0.001
ob5p3 2. 0. J.c 7.2 12.2 12.0 18.8 19.2 9. 0.011 0.051 0.001
04504 1. 0. 4.0 7.0 11.1 50.7 13. 0.012 0.032 0.0
06505 4. 2. 3.0 7.0 9. 40.5 9. 0.166 0.003 0.001
Co510 4. 1. 4.0 7.0 10.2 10.5 44.6 44.7 13. 0.032 0.003 0.0
(3512 J. 1. J.2 4.2 11.5 11.1 38.0 J8.7 8, 0.01d 0.019 0.003
)651J J. J. J.0 5.5 1.2 9.d 42.9 40.0 8. 0.118 0.034 0.003
065 fu J. 1. 3.0 7.0 u.) 34.9 9. 0.099 0.001 0.0
0651!, 2. 0. 3.0 7.0 u.4 S.5 51.5 48.3 9. 0.132 0.036 0.004
04520 J. 3. 4.0 7.0 12.2 12.5 44.0 42.3 13. 0.017 0.006 0.0
04521 3. 3. 4.0 7.0 J.J 45.6 13. 0.042 0.012 0.0
0452.1 J. 2. 3.6 6.5 10.0 41.2 1,3. 0.037 0.005 0.0
04524 2. 1. 4.0 7.0 d.4 8.9 57.9 51.4 13. 0.35d 0.194 0.014
06525 J. 2. 4.0 7.0 10.d 11.0 39.4 42.6 13. 0.014 0.011 0.001
04530 3. U. 4.0 7.0 11.7 11.4 42.3 41.6 13. 0.009 0.007 0.0
06515 2. 0. 2.0 3.5 9.7 Y.4 39.4 42.0 3. 0.044 0.099 0.006
06-545 2. 1. 3.5 7.0 10.2 10.3 39.4 40.5 13. 0.056 0.01d 0.001
06555 O r4o O ***
07401 J. 2. 4.0 7.0 11.1 19.7 13. 0.023 0.023 0.0
07b02 J. 1. 3.0 8.0 11. 23.0 9. 0.107 0.004 0.0
-0760.1 J. 2. 4.0 7.0 14.0 21.7 1J. 0.01d 0.0 0.0
07604 1. 1. 4.0 7.0 1.4 2.1./ 1J. 0.009 0.013 0.0
07401 J. 2. 4.0 7.0 12.1 23.1 13. 0.017 0.019 0.0
07410 4. J. 4.0 '7.5 12.2 23. 14. 0.011 0.0.16 0.0
07412 " 4. 2. 4.0 11.5 12.2 23.4 14. 0.007 0.0 0.0
07613 J. 1. 3.0 6.5 10. 21.6 9. 0.047 0.029 0.007
07614 4. 1. A.0 7.0 12.2 23.5 13. 0.025 0.007 0.0
07415 2. 1. 4.0 7.0 12.2 22.6 13. 0.018 0.034 0.0.
0701007621
..Sus
07430 2. ' J. 3.0 6.C\ 10.4 ***** 37.7 U. 0.022 0.022 0.003
407411 J. U. 2.5 3.0 10.J 10.4 30.4 J9.1 5. 0.040 0.424 ).02J
076.12 3. 0. 2.4 2.4 10.5 10.5 2S.J 37.6 2. 0.092 0.019 0.002
07634 4. 3. 3.5 7.0 8.7 42.0 1J. 0.117 0.009 0.0
07645 2. 0. 3.0 5.0 6.9 43.7 7. 0.026 0.00d 0.0
07640 J. 2. 3.0 4.0 10.d 41.0 6. 0.031 0.001 0.0
07641 3._ 1. 3.0 5.0 9.d 11.7 44.9 40.6 7. 0.090 0.092 0.005
071142, 2. 0. 5.0 6.8 11,7 25.1 15. 0.044 0.025 0.002
07643 2. 0. 2.5 2.5 11.6 10.2 40.5 41.4 5. 0.019 0.437 0.019
07445 J. 1. 3.0 5.0 10.2 47.0 7. 0.254 0.009 0.0
07650 J. 0. 4.8 4.d 11.0 11.1 33.4 38.7 11. 0.080 0.024 0.002
07E51 2. 0. 3.0 5.0 0.8 9.5 50.5 55.3 7. 0.004 0.079 0.075
07652 J. 3. ,3.0 7.0 d.j.. ***** 3/.6 9. 0.091 0.023 0.007
07053 3. 1. 2.0 3.0 9.4 10.1 41.7 40.7 2. 0.077 0.05b 0.003
07654 2. 0. 2.0 2.0 d.2 8.6 17.4 41.1 1. 0.033 0.441 0.056
0/470 3. 3. 3.0 5.0 8.9 41.7 7. 0.240 0.015 0.001
07671 2. 0. 1.8 2.0 8.9 9.1 40.4 42.7 1. 0.035 0.611 0.077
07672 J. 3. 3.0 6.0 8.9 45.5 s1/ 8. 0.095 0.010 0.005
076/1 2. 0. 2.0 3.0 9.0 9.1 37.5 36.6 2. 0.007 0.666 0.011
07474 2. 2. 2.0 2.5 9.4 8.9 40.8 42:3 2. 0.096 0.442 0.262
07675 J. 1. 3.0 4.0 10.7 10.0 41.2 40.5 8. 0.039 0.027 0.004
J7400 2. 0. 4.0 6.0 9.0 54.5 12. 0.009 0.850 0.060
07465 4. 2. 2.5 3.i U.0 S./ 40.5 37.1 6. 0.044 0.004 0.0
07640 2. 0. 3.0 4.3 6.3 43.7 6. 0.106 0.001 0.003
07491 J. 1. 3.0 5.0 9.d 46.0 7. 0.132 0.011 0.005
370',2 J. 0. 2.0 3.0 d.9 41.1 2. 0.077 0.006 0.001
341
Zfe
600'0 ZE0*0 6EZ*0 't Z'PE t'LC 4'6 81 9'1 '74 'h hL6ll
(00'0 C10'0 0f1.0 't Z'OE t'LE 4'01 L'Ot O'Z C'Z .6 *c FL61t
rev() Lovo 6e7'o *r 6'1C o'R 4'7 '4'1 '7 '4 7/611 1'00'0 ZG0'0 SRZ0 'Z 0.0h P'L O'r n.E 0 r 1L611
rove s00.0 1.60 'E S'hf 9'9E f'R 6'L P'h rt '7 'h et6tt 709'0 h00'0
E00'0 Zt0'0
IWO L61'0
'S
'1.
Cht, S'rn S'Fh
'0 6w SR
S'Z O'Z
S'Z O'Z
'0
'0
'4
V S9htt b96tt
910'0 R10'0 FZh'0 'Z ,9'9Z vez 4'6 h'6 O'r. fZ 'F frifitt
700'0 P00'0 760'0 'r G'Zh. o'n or 'h 7. ,1614
700'.0 S00'0 OhZ0 't 8'1E OZ 0'7 'h 09611 0'0 Lt0'0- 170'0 !I. 8'Zh h'Pt l"6 O'S ff 0 1- 50601 Love PAE0 11.0'0 't 8'7h h'Pt I'S 106 3'Z O'Z '0 'E runt PS0'0 440'0 007'0 't t'9F S'L 6'1. 0'7 -o'r '0 'h 70601 0'0 rtvo PE0'0 '8 ,6'Sh 0,1.'6 3'9 n'r "0 '7 10601
960'0 41'1'0 Ghl'O 'Z S'hh 9'LZ L'6 0'01 S'7 7'Z '0 '7 Off60 Z40'0 808'0 1E0'0 '4 A'ff q'tlf cl'Ot L'04 O'f P'7 '0 '7 4060 600'0 P67'0 0E0'0 L.Lt POI vot o'r o'r *0 *r. 4060 040'0 660 57.0*0 'S Vth P9 VII VA VI' 8'7 1 '7 0(1860
200.0 rntro
6h0'0 ;Tv°
SZO*0 huvo
'9 6'9h VP, vfir F'PC
f'!1 l'Zt C'Ft f'Zt
S'E O'f '0
*t
7 ,'r
h436O CC060 74360
rocp 970'0 140'0 '4 r4 (rs vit 69 yr yr 't *7 1Sq60 '100'0 700'0 610'0 'P 'RE 1'71 0'9 O'F '4 'S 04960 0(4'0 eovo um' q'tt vrt crt o'S C'F 0 7 Fh860 hq1*0 66C0 toro '9 S'9k 6'hh t'tt (Li O'h O'f '1 'F Z060 I.10'0 LOCO L'OE hfh f'41 O'Z O'Z '0 'C thP60 154'0 OP0'0 SPO*0 '9 O'h *o* r 0060 191'0 17h'0 901'0 't 9'17 6'P 0'6 of O'Z *7 'C SEPKO
670'0 960'0 6h7'0 '4 r'6h C'04 L'P coR 8'7 4'7 '0 'C bfF160
SLO'0 8W0 11.0'0 'ft f'1; 4.4h L'01 011 O'L 7r060\
t.0 [WO LSt0 't R'RF 914 4'01 4'0 3'7 O'Z '0 'F 10160 ER0'0 or(re P70'0 'Z r'on chz fnt c*nt for 0Z 0 7
1 01'0601
sfltD L0,0 LLO*0 'Zl hth fru, 0'6 h'6 0'9 L'F 'Z 'F (JR6V Oot PrO tern 1 gLE 4'0 6(1 0r 0t 0 I' 1,70601
L64'0 9LE0 010'0 thh h'Zh L'8 4'8 a? z t */: 17060 F90'0 Pt8'0 P00*0 '7I L'ZG 6'94 0*01 7'01 0'9 O'h '0 1' tZ960 h10'O 070'0 194'0 't 671 O'Z O'F. '0 7 ,?,71160
f00*0 E01') cthO*0 S hFh Z'Sh 4'01 1'01 0E 0*E 0 *Z 61R60 010'0 Lt0'0 ZPO*0 'L L'6f 6Ph Ult 7'6 O'S O'f '0 '7 ht060 600'0 811'0 401'0 't USE tail f6 O'Z 0'7 0 7. 11060 hu.-0 6Lh'0 1,10'0 '9 0'61 t'tif 7'11 1!11 IE 3'7 'r rte6o 691'0- h94'0 080'0 '9 h'6f l'ir P'01 9'01 S'E O'E '0 'f 01060 849'0 E670 720'0 '4 l'Sh h'6h hP h8 4'7 S'Z '0 '7 hOR80 p-00 fah0 900'0 't 0-7; clz '0 '7 C08110 .9P7.'0 b600
889'0 bL0*0
h00'0 100'0
'Zt *b
h'CG 8'74 6,,Z 29L
9'P 4'f,
ct*fi
0'9 nZ
O'h 04E
'0
0 '7
*r
70080 10080
870'0 /.97.'0 4/.0'0 ''S 7'04 9*PC 1'6 h'6 f'E 4'7 '0 '7 GfLLO 240*0 ram., PSO*0 ft s.9r hE S'Ot L'6 P*; or 'E 'h 17LL3 Z30'0 8t'10 SO0'0 °I. UP, h'ob 9'9, t'8 O'C '1 'F. OTLLO ta0'0 h00'0 6E1'0 '9 E'9F L'LE h'C1 1'6 O'h O'f 'F StLLO 't100'0
fl'O
770'0 CO0'0
CWO L70'0
"4 'I.
9'7,4 Vit, I'll 7'6 I'll
O'r O'S
O'f O'f
'0
'0
'1
'Z
htLLO fILLO
0'0 400'0 171'0 '9 S'Lh O'h O'f 'h ZILLO zovo 66L'0 100'0 'Z f ?t, O'rh L'L VA V( VZ '7 OILLO 4hO*0 180'0 SO0'0 '9 E'Zh Fhh fE 0 SOLLO b00*0 710'0 IWO 9 tIlf F'Oh UR 0'4 O'r - '1 'F h0/LO tOO'0 P00'0 CO1'0 'I. 7'6f 1.6 R7. 1 *b FGLLO ZOO*0 '150'0 Ot0*0 ft 0*Ott t*hh 1'71 0*L 0 *7 IOLLO LIO*0 66r0 0E0'0 'Z 6'81 7'ff 6't1 1'71 C'F C*7 '0 *7 469LO Ot0'0 060'0 ZWO 'Z l'6E 4'8E 6'6 7'6 6'Z 7'7 't 669/0 Z30'0 S44'0 Z90'0 .1 8'8E 9E C6 .4'6 0'7 O'Z 'C "410
(ET./ (ZT) (TT) (0T) (6) (8) (L) (9) (S) (7) (E) (z) (I)
(PanuT4u0D) aTiVI
17ZE
325
Construction of ,the Industry Data Base
Data on industry, characterigtics were gathered from numerous
sources including the Internal Revenue Service's Corporation Source
Book of Statistics et Income; the Department of Labor's Employment and
Earnings for the United States 1909-1968; the United States Census
ume on "Industry Characteristics"; and the Department of Conmerce's
Ccrsus of Manufacturers a;4 Mineral Industries and the Input-Output
MLtrix for 1958.7
.::ach of these sources were fully investigated and dozens of
industry characteristics were computed before choosing the final set
characteristics included in the regression analysis. AG.in there
was a problem of matching identification codes for at least four
different ceding systems are used in compiling industry informatioh:
the Standard Industrial Classification (SIC), the Standard Enterprise
Classification (SEC), the SEC survey codes, and. the 81 industry
breakdowp usedf_.1 the 1958 Input-Output matrix. It was necessary to
rely 'on industry titleg to merge the several sources of data, again
checking with manpower experts to see that our conversion routine was
7U.S. Department of the Treasury, Internal Revenue Service,
Corporation Source Book of Statistics of Income: 1953 1958, 1961, 1965
(Washington, D.C.: Government Printing Office); U.S. Departmenttf
Labor, Bureau of Labor Statistics, Employment and Earnings for the
United States 1909-1968:(Washington, D.C.: Government Printing Office,
19.68); U.S. Census Bureau, United States Census of Population: 1960,"Industry Characteristics," Series PC(2) 7, 19660J.S. Department of
COmmerce, Census of Manufacturers and Mineral Industries:"1951, 1958,
.1963 (Washington, D.C.: :.7.-Ivernment Printing Office); and Wassily, W.
Leontief, "The Structure of the U.S. Economy," Scientific American,
April 1965.
Ob.
343
I
326'
proper. was made all the more difficult because of changes in
classification Over time even within the same data source. In a.9
good number of cases, for instance; the Statistics of Income had
different codes for apparently the same industry in 1965, 1958, and
1953. Data was collected for several points in time, namely 1953,
1958, 1961, and 1965 in order to obtain "historical" information on
such factors as concentration and after-tax profits. In theory wages
should be related to long -un industry performance, not fluctuations
in such variables. The years chosen reflect different poirts,in the
business cycle in the post-war period with 1958 and 1961 being
recession years and 1953 and 1965 being yeats,of relatively low
unemplOyment.
With the exception of the Source Book of Statistics of Income,
all of these sources are relatively well known to research economists.8
The Corporation Source Book is a set of unpublished'worksheet tables that form the basis for the annuallypublished iepotts, Statistics of Tncome, Corporation Income-Tax-Returns- Beginning with 1942, the Source Book providesdetailed industry group financial statistics by asset sizeclass. No Source Book-was produced for 1952, and for 1962,distribution by asset size Class was provided only for majorindustry groups.
' Assets, liabilities, income, deductions,tax liabilityand distributions to stockholders are provided for about 235industry groups. (These are broken down for each industry)Into 13 to 16 columns providing a total column and sizeclasses ranging from zero assets to assets of $250,000,000 ormore.
8Corporation Source Book of Statistics of Income, "General
Description," op. cit., pp. 1-2.
344
)
327
The Source Book is published annually for returns withaccounting periods ended between July 1 and June 30 of thefolldwing year. Therefore, the 1965 Source Book, for example,includes returns with accounting periods ended betweenJuly 1, 1965 and Salte-lo, abOut .one-half of which arefor the calendar year ending December 31, 1965.
The financial information has increased over the years,but since 1948 relatively few changes have occurred. Thenumber of asset size classes has also changed with datacurrently di-4:ributed by 14 size classes.
Of specite mcern to the researcher is the industry classifi7
cation system usedin the Statistics of Income. It has a number of
shortcomings.9
(
The industrial classification used was revised for 1963to conform with the Standard Enterprise Classification issuedby the Bureau of the midget. The structure qt the StandardEnterprise Classification follows closely along the linesof the Standard Industrial Classification, which was designed asa means of classifying separate establishments rather thanthe companies of which establishments were part.
Year to year comparability of Source Book statistics isaffected by consolidations and mergers* as well as'by changesin the law, the tax forms, awl-the industrial classificationsystems used over the years.
For corporations, industry statistics are greatlyaffected by the returns of the larger companies. Since areturn is classified by industry based on the activityaccounting for the largest percentage of total -:eceipts, thismeans that large corporations with\diversified activities areincluded in only one industry even 'though many of theirbusiness operations are unrelated to\the industry in whichthey are classified. It should therefore be noted in usingthe Source Book, especially for recent \,'ears, that statisticsfor an industry may be either understated by amounts reportedby corporations whose principal activity lies elsewhere, or
overstated by amounts reported by corporations classified inthe industry but having substantial operations in otherindustries.
9Ibid., pp. 2-3.
4
345
328
Unfortunately 'there is nothing that can be done about this
reporting problem and the exact biases are unknown. Attributing
part of the General Motors Corporation's profits from aircraft manu-
facture to the "automobile" industry may tend to either understate
or overstate the profit rate from auto production. For the sake of
present research we had to make the clearly inaccurate assumption
that the profit rate shown by General Motors came from motor vehicle
production alone while the reported profit rate in'the aircraft
industry came only from such primary suppliers as Boeing, Lockheed,
and McDonald-Douglas.
.The industry characteristics finally chosen included the
following:
Market Power Factor ("Concentration") - Most previous research
has used the four-firm concentration ratio to measure the extent of
oligopoly in a given industry. The major problem with this variable
is that it is only measured for manufecturing'industries, leaving
many sectors of economy without an indek of concentration. This is
particularly serious because the real variance in concentration
may be found between sectors of the economy rather than within the
manufacturing sector. We therefore turned to the Source Book to provide
an alternative measure of "market power" which would yield an index
for all sectors of the economy including retail trade, wholesale
trade, and services.
The measure we chose to calculate makes use of the asset size
classification in the Source Book. The "market power factor" (MPF)
346
329
is the percentage of total sales in an industry (i.e. "business
receipts") received hy those firms in the top two asset categories.
jnlike the traditional measure of concentration, the market power
'factor has a variable number of firms. This is apparently a minor
shortcoming for the simple correlation between four-firm concentration
ratios and the market power factor for all manufacturing industries
in 1965 is .89. What is lost in terms of a defined number of firm"s0
is more than gained by the ability to measure "concentration" in the
non-manufacturing sector.
The market power factor was calculated for 119 industries in
the Source Book for eactof three years: 1958, 1961, and 1965. (The
1953 data were not arranged in a manner which allowed calculation of
this measure.) Several weights were attached to the three years and
the several versions of the MPF were tested with.macro data to see
which explained average hourly earnings best. It turned out that
equal weights on each of tht years predicted wages as well as any
other MPF index. This "unweighted" market power factor was then
merged onto the SEO tape.
After-Tax Profit Rate - The profit measure also comes from the
Source Book The index of profits used is computed by dividing "net,
income" into "total assets." Thus the profit rate used is on an
assets b s rather than on sales. This measure is also historical,
using unweighted rates from 1953, 1958, 1961, and 1965. Again the
use of the unweighted measure came from ,,sts on macro industry data.
347
330
Capital/Labor Ratio - The actual measure used in "depreciable
assets" divided by the number of production workers in an industry.
Thil, was a more difficult and somewhat more unreliable measure than
the others for it relied on the merging of Source Book data on
assets with employment figures from Employment and Earnings. Insofar
as the SIC and SEC classifications do not link perfectly, the capital/
labor ratios are imperfectly measured. The data refer only to.the0
year 1965 because of the extremely time consuming process of matching
other years for the two sources.
Government Demand - This was measured as the percentage of total
final demand in each industry purchased by the federal, state, and
local governments combined. It is derived from the 1958 Input-Output
Matrix in a straight-forward manner. Its main drawback is that the
Input-Output industry breakdown is quite different from the SIC and
SEC classifications which poses soma difficult linking problems.
Percent Minority Employment - The final industry variables
used in the regression analysis were computed from the 1960 Censds
volume on "Industry Characteristics." The percentage of black male,
black female, and white female employment in each census industry was
calculated and added to the SE0 file.
All of the industry data is reported in Table A.8.
348
.01016
01017
01011
.02126
02136
02146
02156
03000
04706
04207_sAwmiits.PLANING
04208
04209
TABLE A.8
TBE INDUSTRY
___INDUSTRY
CONC.
AGRICULTURE
0.1021
DATA
PROF.
0.0164
0.0240
0.0059
0.0344.
0.0134
0.0414
0..0243
0.0245
0.0406
0.0406
0.0360
0.0419
K/L
GO
V. D
.
____NA_:_0.0240 0.2067.
NA
-0.1209 0.2155
NA
-0.12°4 0.1165
39.2072_0.1091 3.0420_
19.7765
0.0221 0.0560
23.5(29
0.0
0.1149
25.083170.0010 0.1076
11.0
209.
NA
0.1265
NA
0.0009 0.2747
12.8503
_0.0
009
0.2288.
6.1935
0.0009 0.3172
3.9548
0.0495 0.2439
FORESTRY
FISHEPTES
METAL MINING
COAL. MINING
PETROLEUM C GAS EXTRACTION
NlvmETALIC MININGF._OUARRYING
0.2484
0.0923
_____0.6063
0.3224
0.4425
0.1488
CONSTb,OCTION
tmlnimc,
MILLS__
MISC wnno PPooucis
FURNITURE E FIXTUPES
0.0216
0.2171
0.2371
0.1775
0.0693
04216 GI ASS F GLASS PRODUCTS
0.5577
0.0750
19.3688_0.0011
0.2855_
04217
CFMENT.CONCRETE,GYPSUM,PLASTER
0.2525
0.0528
29.8127
0.0010
0.1583
04218
STRUCTURAL CLAY PRODUCTS
0.5234
0.01°0
13.0643
0.0010
0.2626
_04219
POTTERY E PRATED PRODUCTS
0.2387
Q.0262
4.5705
0.3622
04?36
MISC MINERAL & STnNF PRODUCTS
0.4633
0.0544
_0.0010
18.1580
0.0010
0.2279
04737
BLAST FURNACFS.STEEL WORKSI1)
0.7180
0.0481
31.8519
0.0061
0.1639
04239
PRIMARY NONFERROUS INDUSTRIES,
0.5141
0.0316
69.3857
0.0525
0.2006
04246
CHTLERY.HAND TnOLS,HARDWARE
0.1581
0.0754
6.0327
MA
0.3280
04247
FABRICATED STRUCTUhAL METAL PRODUCTS 0.1044
0.020?
6.7429
0.0
0.1402
_0424,1MISC.FABRICATEO METAL PRODUCTS._____ _0.2372
0.0512
11.1692_0.0257
0.2370
04256
FARM MACHINERY C EQUIP
0.7339
0.0348
17.2499
0.00970.476
04257
OFFICF,COMPUTING.ACCTING MACHINERY
0.7267
0.0741
34.6225
0.0737
0.2534
04258
MISC MACHINERY
0.0545
9.3115_0.0502
0.1520
04259
ELECTRICAL MACHEOUIP.SUPPLIES
_0.1816
0.5204
0.0553
7.8971
0.0973
0.3735
04267
mnTnR vEHICALS.mOT VEH EQUIP
0.8786
0.770
23.4105
0.0332
0.2053
04261
AIRCRAFT C PARTS
0.8406
0.0426
11.4435
0.5137
0.1931
04269
SHIP C 9OAT BUILDING F. REPAIRING
0.3928
0.0149
3.8691
0.1865
0.1546
04276
RAILROAD F.
MISC TRANSPORTATION EQUIP 0.4538
0.0289
14.4807
0.1865
0.1627
_042862ROFESSIONAL
EQUIP C SUPPLIES
0.2547
0.0557
9.3756
0.0972
0.3210
04287
PHOTOGPAPIIIC EQUIP
F. SUPPLIES
0.6679
0.0107
37,1984
0.0972
0.2891
04287
WATCHES E CLOCKSI2)
0.61)5
0.0339
4.0E
15-2
--NA
.0.5313
.05306
MEAT PRODUCTS
0.3402
0.0235
8.8045
0.0072
0.3639
05307
DAIRY PRODUCTS
0.4055
0.0481
21.7709
0.0072
0.1635
05308 CANNING C PRFSPPVING-FRUIT C VEG
0.2748
0.0427
9.2858
0.0072
0.4929
.05309_GRAIN-m111
PRODUCTS
.0.3866.0.0525
27.9480_0.0072
0.2288.
05316 BAKERY PRODUCTS
0.2506
0.0538
11.0182
0.0072
0.3278
05317 CONFECTIONARY C RELATED
PRODUCTS
0.2460
0.0765
5.7786
0.0072
0.5652
115.
31.1
.14E
VE
RA
5LIN
DU
STR
IES
0.3492
0.0472
__19
16ji
IA...
.0.0
072_
0.
1866
.
cir
Ino
TABLE A.8 (Continued)
INDUSTRY
05319 MISC FOOD PREPARATIONS
o5,24-rnou:cn mANUFACTORES
05346 KNTTTING
(15147 DYEING C'FINIcHING TEXTILES(3)
05348 FLOOR CriVERINGIEXC HARD SURFACE)
05349 YARI.THREAD C FABRIC MILLS
05356 MISC TEXTILE MILL PRMUCTS
_
05 366 t""ARFL E ACCESSDRIES
05367 MISC FABRICATED TEXTILE PRODUCTS
05386 DIJt.Q.PA ()FR
C PAPERBOARD MILLS
05387 PAPFPBDARD CONTAINERS 1
BOXES
05389 MISC PAPER C PULP PRODUCTS
05196 NEWSPAPER PuPLIcHING C PRINTING
05398 PFINTING.PUBLISHING.ETC(4)
05406 SYNTHPTIC FIBERS
05407 DPuGS E mFDICINES
05408 PAINTS.VARNISHES C RELATED PRODUCTS
05409 MISC CHEMICALS E ALLIED PRODUCTS
05416 PFT:4111FuM REFINING
05419 VT SC PETPDLEUm E COALORODUCTS
05476 RUBBER PPOOUCTS
05429 MISC PLASTIC PRODUCTS
05436 LFATHER-TAINED.FINISHED.CURRYED
05437 FOOTWFAR.EXCEPT RUBBER
'
_05438 LEATHER PRonucTs.Fxc FOOTWEAl?
0(2,506 RR C RA1LwAY EXPRESS SERVICE
06507 STREET RAILWAYS $ BUS LINES
06508 TAXICAB SrPV10E
06509 TRCCKING SERVICE
06516 wARPHDUSING r, SMRAGE
06517 WATER TRANSPORTATION
06518 AIR TRANSPORTATION
06519 PETROIrUu E GASOLINE PIPELINE
_06526 SFRvICES INCIDENTAL' TO TRANSpORIL.
07536 RADIO PPDAOCASTING C T.V.
01538 TELFPI-ONF(w1RF C RADIO)
_07511_TFLEGRAPHIwIRE E ROM)
08567 ELECTRIC LIGPT C PEIWFR
08568 GAS
C.STFAM SUPPLY SYSTEMS
.08569.FLECTRIC-GAS UTILITIES.
---.
CONC.
PROF..
Grytii. THIN_
------
0.3525 0:0467 _25.0241_
0.8967 0.0138
12.2338
0.1323 0.0382
4.0431
0.1744 0.0234_ 3.6054_
0.3168 0.0241
10.5136
0.3871 0:0345
33.4269
0.3947 0.0301.
2.4748
0.0449 0.0283
1.1906
0.1043 0.0247
2.2997
0.5858 0.0504
17.4776
0.1559 0.0556
8.1469
0.2147 0.0638
12.1658
.0.2117 0.0666
14.8525
0.0929 0.0478
9.3934
0.7331 0.0881
52.6891
0.6239 0.0993
32.4553
0.3819 0.0624
25.3242
0.5747 0.0572. 50.7991
0.9347 0,0320.510.8074
0.3071 0.0422
26.8645
0.6173 0.0522
17.1447
0.0072_0.3326
0.0
0.6173
0.0024 0.6976
0.0024 01.2761.
0.0024 '0.3845
0.0062 0.4470
0.0458 0.3890
0.0091 0.7753
0.0'58 0,6072
0.0089 0.1839
0.0010 0.3359
0.0089 0.3878
0.0213 0.2056
0.0213 0.369
0.0011 0.2633
q.0471 0.3142
0.0012 0.2364
0.0826 0.2287
0.0641 0.1497
0.0641 0.2750
0.0282 0.3063
.0.0212 0.3937_
0.0
0.2174
0.0080 0.5543
0.0080 0.5577.
NA
0.1370
NA
0.2139
NA _0.2163
NA
0.1246
NA
0.2969
NA
0.1980
NA
0.2556
NA
0.3/91
NA
0.3122
0.0
0.2709
0.0402 0.5737
0.0402_9.4325_
0.0415 0.1687
0.0415 0.1896
_9.0415 0.1687
0.0960 0.0407
5.3437,
0.0873 0.0281
9.4909
0.2468 0.0375
2.1882
0.0179 0.0317 _3.5414_
0.9257 0.0095
NA
0.1476 0.0140
22.7297
0.2833 0.0429
NA
0.0275 0.0384
8.1359
0.9275 0.0384
91.1336
0.2178 0.0160
NA
0.7105 0.0191
NA
0.3750 0.0373 217.5037
.0.1613 0.0216._
NA
0.5348 0.0622
16.9890
0.9590 0.0431
75.9472
0.7835 0.0264
38.3041
0.9442 0.0200 162.8866
0.8295 0.0229 169.9157
0040T 0.0268 290.7214
10
TABLE A.8 (Continued)
_INDUSTRY..
- _
08576 WATER SUPPLY
08578 SANITARY SERVICES
C9606 MOTOR VEHICLES C EQUIPMENT
_09607..PRUGS CHFmICALS E ALLIED PRODUCTS_
CoNC._PRoF2___K/L
0.1402 0.0194_71.5791
0.0415 0.2031
0.1402 0.0174
71.5791
0.04150.2980
0.3350 0.0433
2.6735
0.0090 0.190
0.2774 0.0565_2.7972
0.0000 0.3019
01608 DPY nonOs C APPAREL
0.0349
0.0233
2.0838
0.0090
0.3559
09609 Fnon F. RTiATED PRODUCTS
0.0146
0.0293
3.7193
0.0090
0.2879
09616 rAP4 PRODUCTS -RAW MATERIALS
0.2198
0.07,07.
NA
0.0090
0.2213
09617 riFCTPTCAL OrnnS n HARDWARE1-51
0.0903
0.3318-2.8999
0.00y0
0.2440
00618 mtCHINFRY.TQUIPFENT C SUPPLIES
0.0590
0.0391
2.4919
NA
0.2110
_10636 Font) STORFS1ExC DAIRY pRonuCTst_.
0.0496
3.4499
0.3685.
10637 DAIRY PRODUCTS C MILK RETAILINC
0.3882
0.0496
__.0.0090
NA
0.0090
0.3685
10638 GENERAL MERCHANDISE RETAILING
0.4285
0.0378
4.1517
0.0 '390
0.6920
_10619 LIMITED PRICE VARIETY STORES
0.6344
0.0436
0.0 90
0.8130
10646 APPAREL
E ACCESSUR IFS STORES
0.0668
0.0274_3.9344
2.4320
0.0u>)
0.e919
10048 EuRNITUIE C FURNISHINGS STORES
0.0632
0.0126
3.5125
0.0090
0-0,019
_J06,. HnusFunin ApPLIAmCES,T.v. E RA0.10
0.0632
0.0126
NA
0.0090
0.2717
10656 mOT08,VFHICLE C ACCESSORIES
0.0198
0.0133
2.8805
0.0090
0.1778
10657 GAS0LINE SFRVICE STATIONS
0.0195
0.0328
NA
0.0090
0.1096
_10E58 nPun STiRgs
10(5; EATING C DRINKING-
_ _ _
PLACES
0.1505
0.0573
0.0482_1.9740_ ...0.0090
0.0194
1.6875
0.0090
0.5399
0.6358
10666 HA8D,AARE E FARM EQUIP STORES
0.0375
0.0130
0.8246
0.0090
0.2182
.10676_044FR): BuILDINA.mATERIALS_
Cs.0427
0.0254
0.0090
0.2043
10678 LIQUOR STORES
0.1220
0.0260
_2.5258.
NA
0.0090
0.2622
10686fiFWELRY STORES
0.0518
0.0184
NA
0.0090
0.4470
_11706 BANKING
F.CREDIT. AGENCIES
0.5180
0.0046
1.0236_ 2,00710.58-44
6.3189--
11716 SHUR r. CIMMOD BROKERAGE CO (6)
0.0649
0.0136
1.5902
NA
11726 INSURANCE
0.7592
0.0115
NA
0.0071
0.4651
_11136 REAL ESTATE (7)
0,.0332
0.0127
NA
0.3957
12806 ADVERTISING
6.1818
0.0531
_0.0061
NA
(1.'04620.3954
12807 MISC BUSINESS SERVICES
0.0560
0.0377
NA
0.0462
0.4105
1780.11_NYo ,r,Fr!AtR SERVICES E GARAGES
0.0865
0.0232
NA
0.0762.0.1606
12809 MISC REPAIR SERVICES
0.0620
0.0398
NA
0.0271
0.1243.
13926 HOTEL
E LODGING PLACES
0.1337
0.0031
11.4702
0.0271
0.6004
_13826_LAUNDERTNG,CLEANING C DYEING
0.0747
0.0338
3.0no
0.0271
0.6433.
13836 SHOE REPAIR SHOPS
0.0620
0.0398
NA
0.0271
0.2763
13838 RARnER
F.BEAUTY SHOPS
0.0355
0.0505
NA
0.0271
0.6110
_14846 THEATERS
F. MOTION PICTURES
0.1062
0.0132
NA___ 0.0
0.3846
14848 pnwuNC AiLf7,6000l C BILLIARDS
0.0473
0.0169
NA
0.0
1481,9 MISC ENTERTAINMENT F RECREATION
'0.0473
0.0169
NA
-0.0050
0.4045
APPENDIX B
MEANS, STANDARD DEVIATIONS, AND CORRELATION MATRICES
352
VARIADLE
SCHCOL 1
WAGE 2
UN MEM 3
MIG<50 4
S.AT16 S
EXP' 6
TRAIN 7
PHY DM 8
NEG WT 9
GED- 10
.3VP 11
UMOCC 12
%iFCCC 13
.caocc 14
DA/PW 15
NYL/PW 16
AVEMPF 17
FUSL 16
ITSLEX 19
ATPEAS 20
MININD 21
EDS16 22
UNXMPF 23
%FMOCC 24MINOCC 25
335
Occupation Stratum 1-3 White Males
MEAL!
9.0507246382.707318841
0.57971014490.47101449280.253623188429.63768116
0.9420289855D-013.2971014491.3115942031.9521739132.292028986
0.15378985510.14364492750.3139130435D-01
STANDARD DEVIATION
2.8369883890.87618411210.49540360610.50097756970.436669448612.00604594
.0.29317494740.67714967550.76225486320.15341246880.5531675997048134619654D-010.19423951660.6059900926D-01
24.527083603.032517139
49.003754873.687716972
0.3629292271 44
0.4064332987D -01 0.6818532,97D-010.66475136461 -01 0.8142093 52D-010.41674e3766D-01 0.1734286728D-010.3012124489 0.16325534051.956521739 3.7569780630.461270531 0.30737459640.1750362319 0.23606847980.3288260870 0.2045273673
35a
CORRELATION MATRIX
VAR/ABLE
RUBBER
12
34
56
7
11.00000
0.248569
0.568278D-01
-0.109377
-0.275604
-0.476702
0.310148
-0.102892
21.00000
0.282418
-0.141940
-0.350388
-0.231745
0.189954
0.6888600-02
31.00000
- 0.788542D-01 -0.212231
0.6118280-02
-0.7720580-01
0.703220D-01
41.00000
-0.4956590-01
0.100180
-0.155215
0.100879
51.00000
-0.4620760-02
-0.1693970-01
0.1484710-01
61.00000
-0.355209
0.3219180-01
71.00000
-0.105241
81.00000
VARIABLE
910
11
12
13
14
15
16
iUBEIER
1-0.951219D-01
0.7605320-01
0.8630670-01
-0.609657D-01
0.6071280-01
-0.4066340-01
-0.165087
-0.4096430-01
2-0.126938
-0.3984180-71
0.8906610-01
-0.9550010-01
-0.193301
- 0.272B72
0.1809680-01
0.137444
30.2073020-01
-0.2630710-01
-0.2030490-01
-0.8516380-01
0.3674450-01
-0.11653d
0.4993440-01
0.262830
40.5249540-01
-0.3716340-01
-0.5746930-01
0.103286
-0.7245440-01
-0.4795090-01
-0.2809490-01
-0.3903520-01
50.2399520-01
0.6253340-01
0.446924D-01
0.473355D-01
-0.3567710-01
-0.448784D-01
-0.3973280-01
-0.130759
6-0.1229910-01
-0.9080330-02
-0.8363710-01
0.475820D-01
-0.122128
0.4555370-01
0.8251410-01
0.878570D-01
7-0.132308
0.133360
0.171196
-0.677227D -01
0.2481740-01
-0.106858
-0.5981570-01
0.2770470-02
80.696109
-0.740524
-0.728280
0.824973
-0.558752
-0.478152
0.192873
0.10988t
1.00000
-0.701810
-0.546288
0.683003
-0.167309
0.107956
0.5127750-01
-0.6897480-01
10
1.00000
0.737764
-0.675669
0.29104/
0.1511508
-0.187357
-0.5251440-01
1I
1.00000
-0.713320
0.139367
0.104047
-0.231765
-0.9494020-01
12
1.00000
-0.570883
-4250883
0.180390
0.1955770-01
1314
1.0000
0.600590
T1.00000
-0.157098
-0.149111
-0.1n0514
-0.232338
15
1.00000
0.766573
16
1.00000
L.)
'1/4
...1
C.7%
VARIABLE
NUMBER
17
18
19
20
21
22
'
23
10.141655
-0.5223770-02
0.4286880 -C1
0.149627
-0.6506290-01
0.1253540-01
0.7897730-01
0.3951680-01
20.275904
0.8780920-01
0.127912
0.263330
-0.381537
-0.280848
0.319694
-0.229096
30.273440
0.101074
0.115482
0.323286
-0.159091
-0.161839
0.689850
0.3184020-03
4-0.1034050-01
-0.2695420-01
-0.3071950-01
-0.9623670-01
-0.2664350-01
-0.8599330-01
-0.7296330-01
-0.7192530-01
5-0.152285
0.190044
0.207108
-0.128781
0.1938000-01
0.896623
-0.226826
-0.4087580-01
6-0.3509860-01
-0.3793360-01
-0.112249
-0.1449090-01
0.6247850-01
-0.146316
0.3041000-01
-0.8879450-01
70.8894890-01
-0.7855891 -01
-0.4844580-01
0.3263060-01
-0.9423040-01
0.7001520-01
-0.1695480-01
-0.7010470-02
a0.145505
-0.109526
-0.107072
0.5882310-01
-0.404059
-0.5513790-01
0.7874310-01
-0.582489
9-0.6525230-01
-0.182528
-0.206977
-0.102871
-0.4154.690-01
-0.2072310-01
-0.2603830-01
-0.109951
10
-0.4711090-01
0.204042
0.229803
0.3227020-01
0.198341
0.133140
-0.1165660-01
0.280161
11
-0.2350870-01
0.263119
0.275054
0.3880520-01
0.8271320-01
0.8588190-01
0.1012360-01
0.141381,
12
-0.3354200-01
-0.188988
-0.241124
-0.163714
-0.184729
0.9833900-02
-0.4489130-01
-0.534130
13
-0.124535
-0.9581650-01
-0.5314170-01
-0.3785580-01
0.446750
0.3443690-01
-0.6286950-01
0.979036
14
-0.249226
-0.108167
-0.160696
-0.281522
0.68'370
-0.9767410-02-0.196532
0.757455
15
0.297717
0.8908200-02
-0.5022800-02
-0.104221
-0.173451
-0.6675610-01
0.277756
-0.167538
16
0.627292
0.9947420-01
0.146574
0.417685
-0.305693
-0.128465
0.605029
-0.142350
17
1.00000
0.142604
0.219425
0.487101
-0.370070
-0.110573
0.762784
-0.166445
18
1.00000
0.947289
-0.9668490-02
-0.230682
0.180700
0.162758
-0.106605
19
1.00000
0.116404
-0.282053
0.223627
0.201473
-0.8497650-01
20
1.00000
-0.375978
-0.102924
0.446164
-0.103415
.21
1.00000
0.5846090-01
-0.320200
0.524530
22
1.00000
-0.183742
0.2580300-01
23
1.00000
-0.102209
24
1.00000
VAMIAULE
23
SUMBER
10.213631D-01
2-0.302409
3-0.3350450-01
4-0.4193730-01
5-0.2835280-.01
6-0.8356320-01
7-0.3502680-01
a-0.344233
90.144742
10
0.546339D-01
11
-0.720523
12
-0.218773
13
0:902962
14
0.774483
15
-0.121629
16
-0.156524
17
-0.205454
18
-0.198211
19
-0.193983
20
-0.184477
21
0.531948
22
0.336934D-01
23
-0.135826
2*
0.941776
25
1.00000
0 0 rt
338
Occupation Stratum 1-3 Black Males
VARIABLE
scHrot.
m E A14.
7.632411067
STANDARD DEVIATION
3.405652531
WAGE 2 2.16'94C7115 0.8149864818
UNInN 3. `0.4743083004 0.500329?723
MIG<50 4 0.4584980237 0.4992622702
S. 16 5 0.7984189723 0.4019761088
EXP 6 33.10276680 12.64649523
TRAIN 7 0.94861660080-01. 0.2936045834
PHY OM 8 3:470355731 .0.6577609567.
MCG WT - 9 1.3813992C9 0.8493935201
SVP 10. 2.088932806 0.45.76033661
%OvilCr. 11 0.2539683794 0.1793019063
ZWFOCC 12 0.1003873518 0e1630923884
2:8FOCC 13 0.3325691'00D-01 0.64527350560-01
DA/PW 14 )6459372127. 33.26334179
AVFmPF IS 0.3519994730 0.2770429297
FDSL 16 0.32050215280 -01 0.73419058830-01
ATPrAS" 17 0.39'56719368D-01 0.19102449120-01
ZMIN1N IR 0.3175426489 0'.1715143466
EDSI6 19 5.806324111 4.232109816
UNXOPF 20 0.2322388669 0.3108318211
7,FMOCC 21 0.1336442688 0.2155'543424
ItMINOC 22 0.38761264E2 0.2119432580
356
C0R9EkAT1CJ mA73IX-
3,171341.E.
4000F9
I4,
2:6
li, 4" S 6
1
1.00000
41`,
2
0.176282
1.00000
C'
3 \
0;121358
0.496698
1.00000
4
-0.140870
-0.181349
-0.141208
1.00
000
5
-0.210970
-0.338/2.6
-0.252756
-0.111057
1.00000
6
-0.654943
-0.5499590102
-0.3595590-01
-0.9925220-01
0.132110
1.00000
7
0.265190
0.180873
0.4367000-01
-0.8132070-01
-0.5448790-02
-0.227068
8
- 0.159886
0.124134
0.103114
0.114056
0.4484690.01
,0.8719020-01
16
1.0000:
-0.8812070 -01
8
s 1.00000
9ii7I401.0
910
11
12
13
14
15
r6
603/R
1 2 1 4 5 6 7 8
10'
1112I3
14 15
16
- ('.7816410 -01
4.122912
-0.167566
0.150160
0.169774
-0.155645
0.3063910-01
0.9235640-02
4.7591211-11 -0.47994809 .1 -0.212127
-0.41r9170-01 -0.142056
0.155235
0.535418
0 0 39909
0.4661410-31 -1.3591130-01 -C.101843
=0.2(93740-01 -0.5914610-01 -0.116395
0.445357
40.3060210 -01
0.491010-01 -1.9060170-01
0.6443830-01 -0.8698791-01 -0.57e1170-01
-0.5775550-01 - 0.3268900 -01 -0.4518080-01
n C
lli(?4
..., .
0 7.
ti
0...1405150-01 -1.6179420-01
0.2606340-01
0.6069600-01
0.4071030-01
0002113050 -02 -0.156224.
0.2502940-01'
g
0.4075300-02 -0.:43852D-01 ,0.396385D-01 -2.4338730-01 -0.7577040-01
0.172505
0.3058360-01
0.3005450-01
- 0.143465
0.1270450-11
0.649r6,0-01
0.143465
0.173288
0.106997
--0.3424640-01
0.5419790-01 -0.531-7780-02
0.634806
-1.606213
0.4561740-01 -0.615905
- 0.629007
0.325216
0.5930070-01
'
..00000
-0.346672
0.37494-
-0.327993
.-:6323710-01
0.2310180-01
0 290238
.4 0.157582
0.530013D-01
(1) rt 1 a
1.90000
-0.198680
0.309674
1.00000
-0.454323
-0.270263
-0.211116
-0.261278D-01
.....
C.JO
-:;;;7g
,1.3
-0.109907
-0.102988
-0.9985640-01
i W1.00000
0.745956
- 0.186105
-0.197400
-0.540830-01
01.00000
-0.224251
-0.209997
-0.482009D-01
vAP148LE
17
140041-4
f -
C.10c797
18
1.131158
19
0.404111
20
0.9260329-01
210.393596
-0.324944
-0.191594
0.561715
30.407791
-1.205,162
.-0.119489
0.788142
40.5145800-01
-0.116212
-0.194444
-0.117332
..4
5-0.5459700-0i
0.3919220-01
C.6,.00739
-0.708562
6-0.5091501-01
-0.7317680-01
-0.334903
0.1199910-01
7C.129470
-1.1940460-01
0.168137
0.5534440-01
60.217897
-1.424305
f-0.6172990-01
0.229329
9C.109411
-1.361913
1
0.1571910-01
0.134127
10
-0.10507
0.241110'
0.1551070-01
-0.114688
11
I.C.136496
1.175399
-0.8990270-01
-0.120788
12
-0.9789410-01
0.474112
0.160317
-0.116263
13
-0.161396
.Y.602463
0.158492
-0.1:7699
14
.6196645
-0 527499
-0.9565470-01
0.5457960-01
IS
0.49558.5
- 0.411127
-0.119964
0.776040
16
-0.124086
-1.151605
0.2729900-01
0.P920720-03
r1.00000
-3.309779
0.1153210-01
0.528674
18
1.00000
0.128388
-0.354616
19
1.00000
-0.11721^3
1,.00000
21
-72
,
1.00000
1.00000
0.356132
-0.2552700-01
co
1.00000
0.8123480-01
w
21
22
-0.101881
-0.285108
4`Y1
7,:c
w ....
0.164443
0.2971480-01
-0.3890360-01 -0.132492
-0.831768.A-01 -0.2974140 -01
0.5811060-01
0.8115010-01
-0.5550990 -01 -0.2376790-01
0:124312
0.6077060-02
,
-0.654326
-0.626883
-0.4093070-01 -0.334909
0.323468
0.160698
e.0.43588^
0.402604
0.979924
0.6/0576
0.865760
0.622387
-0.207141
-0.304464
- 0.736169
-0.327319
-0.5590570-01 -0.141336
-0.13176s.
0.540646
0.168864
-0.156794.
1.00000
-0.248459
0.698243
0.9568410-01
-0.261143
0.648288
1.00
000
J
340
Occupatiiin Stratum 1-3 White Females
kt.411AnLF
SCHOOLwArrE 2
UN ',FM 3
1q1C,<50ssaT16 5
FXP 6rwmpire n4 R
mFAN
9.()307692311.739646154
0.32307692310,5076923077.415104615433.23076123
0.3076n23077D-012.723076923
N1t wT 9 1.200000000GFO In 1.901538462"VP 11 2.3/5384615T,rl'AnCr 1? 0./510769211D-01ZwrOCC 13 0.4314000000l'AFrICC 14 0.1057692308.rdi/PW IS 11.074616474YL/PW 16 1.950009438AvrmPF 17 0.2597128205FOSt. In 0.25216372331)-61Fnst.f-'x 19 0.40577638370-01ATPPAS -20 0:40669230770-01MTNINO 71 0.4629771389E0S1( 27 3,353846154ONV,TF ?3 0.1493641026D-017,Fm0CC- 74 0.54416923P1M1'19 0C 75 0.6292769231
358
STANDARD DEVIATION
2.5554429090.46221329660.4712911888'0.50383147370,.496623212710.6576.9855
0.1740560530.64970407460.66614562970.7683/648973D-0140.45264888800.62124407180-010.22463940310.1?'79L49827124528014582.1700780455
048960114730./3392743295D-01.0j:545927902-2D-010.16366896120-01/0.17656814344.431530383
/0.18279561031 0.26196156.500.2201675859
CnaPcLartry mArDix
Vey
lnA
kc1
23
45
67
8O
N0).8E9
0 0
11.00000
0.451280
0.277039
-0.6086570-01
-0.318027
-0.569956
0.6810400-01
0.193433
21.00000
0.390739
- 0.121867
-0.6040770-01
-0.250028
0.158314
0.381390
(11
31.00000
0.2227180-01
-0.248545
-0.105287
0.6749720-01
0.245723
41.00000
-0.106639
0.7386550-01
-0.2741470r02
-0.8885690-01
0S
1.00000
0.9378560-01
-0.150188
-0.2533050-01
61.00000
-0.8812790-01
"1.00000.
-0.157610
Lei
7--,
0.7653410-01
8. \
-\
1.00000
VA91%a1f
imunFR
9in
11
12
li
14
15
16
rt C a
C"
I-P.2202110-11 -1.141100
-1.1)1121
-0.3242040-01
0.7605911-01 -0.171614
-0.21976:
-0.1559790-01 -0.5700460-02
0.294128
CIT
7-
C.)
102
ii,%
-000
5186
P-3
P-1.1/004
-0.217544
1.1146540-01 -1.I177o7
-0.,011046
0.8151110-01 -0.3511780-01 -0.220865
0.149524
0.9483670-01
0.249074
rt
4-0.7448799-01
1.3725690-02 -0.7545790-01 -0.133562
-0.215965
0.134298
0.153374
0.248288
0.203656
5-0.1899230-31 -9.4157730-01 -0.9833640-01
0.4866530-01
0.121683
-0.150719
0.8809430-01 -0.2258460-01
.,,
-0.200776
9.3581260-01
0.4651970-01 -0.1326560-01
0.7679660-01
0.143123
-0.2722190-01 -0.108715
70.898650-11 .0.4314310-01
0.7320250-01 -0.3201610-02
0.4018330-01
0.1025820-02
0.217408
0.144243
ft
0.166071
-0.542203
-0.569470
0.663882
-0.529709
-0.523451
0.9872320 -01
0.118=
91.00030
-1.354113
-0.2589810-01
0.540518
-0.300892
0.284041
-0.9310510-01 -0.9702190-01
11
1.00000
0.421421
- 0.672165
0.190967
.-0.185719
11
1.00000
-0.368233
0.407443
-0.116259
0.109440
0.3375280-01 -0.4887940-01 -0.5939920-01
-0.78'51
17
-0.7719800 -03
-0.149887
-0.124749
13
1.00000
0.5533480-01
0.218605
1.00000
0.136653
14 35
1.00000
-0.284549
-0.369998
16
1.00000
0.823977
1.00000
.P I-
e Ca cr)
v4411911
17
tq
19
20
21
22
Avn$4f4
I6.271762
A.11n41,
1.191157
0.739204
-0.231611
0.2661850-01
,1w-550.66-6i
9.133107
0.407/19
0.120011
0.45
4115
-1.556114
A0.1''"14!:
0.1''"14!:
,-1.146810-11 -0.1169470-01
0.2351'71
-0.272806
-0.137886
4.4
iG47
83-1
1- 1.1P8157
- 0.22)072
-0.123324
00.2199580-01 -0.172692
50.161118
-1.162115
-0.112125
-0.3317381-01 -0.3839780-01
0.904828
6-0.2797440-61 -1.176954
-0.246495
-0.416369
0.110637
-0.7685400-01
70.613156n-1i -1.115354
-0.9367870-01 -0.259974)-01 -0.3113470-01 -0.135894
80.1579/7
1.1599760-01
0.214485
0.416494
-0.366346
0.4541890-01
9-n.)49534
-1.7571710-01 -0.7305300-01 - 0.7823260 -01
0.307002
-0.1905460-01
10
-0.1
9744
41.859/510-81 -0.5935490-02 -0.196272
11
*-0.342506
1.73735?
0.5723930-01 -0.73/671
0.154775
--'0.100737
12
-1.1694610-52 -1.4314463-01
0.1620050-03
0.177114
0.3318780-01
0.5871430-01
13
0.115229
-1.141559
-0.261581
-0.177093
0.4459800-01
0.108633
14
-g:4=
-1.4822920-01 -0.750746
-0.653086
0.485321
-0.203332
IS
-0.276182
-0.883140-0/
0.305163
-0.438635
0.9085420-01
16
0.545518
-0.104143
0.129237
0.562.02
-0.409843
-0.1656340-01
17
1.00110
-0.104613
0.183446
0.535479
-0.507536
14
'
1.e0000
0.857257
0.175314
-0.139882
0.249562
-0.122891
19
1.00000
0.397594
-0.266499
-0.3347750-01
70
-0.623418
0.124842
71
1.00000
1.00000
-0.154003
77
1.00000
23
1.00000
74
.
-0.100739
1.00000
23
0.328252
0.419151
0.75/618
0.8311980-01
-0.6407680-01
-0.156543
0.189009
0.191578
-0.150410
-0.19/066
-0.272358
0.1513550-0!
0.2171110-01
-0.243927
0.409904
0.262693
,-0.6504950-01
0.505696
-0.6771860-01
-0.1372810-01
0.358497
-0.427783
0.7450020-01
24
-0.252509
-0.354991
-0.135868
0.188518
0.2935700-01
0.134862
0.3449910-01
-0.703774
-0.115882
0.437976
0.109076
-0.737656
0.873096
0.535114
0.460118D-01
-0.123562
-0.314227
-0.343756
-0.377601
0.274714
-0.7374350-02
'
C n rt O to
C) ft C B
VAQ1AriLF
75
NO14":11
1-0.778941
7-0.415044
1-0
.14
0.18o617
So.4465160-01
56.151076
70.461.0-11
9P.14610211 -ni
Inn.
3446
416.7 5 p7740-01
12 11 14 is
-C.595515
0.51
7212
0.594400 .
0.i964690 -11
15
-0.991.4050-31
17-0
.147
4cS
1.1
-P.1
461;
"719
-n.4
0a76
779
-C.4
1341
121
0.315277
22
9.7793140-02
21-6
.115
492
24
6.981695
2S1.
nono
n
0 0 ft
C.)
343
Occupation Stratum 1-3 Black Females
VARTABLP MEAN STANDARD DEVIATION
sr7v101 1 1081791825 3.014919250wAnr 2 1.158467153 0.5010061795woinm 3 0.2919708029 0.4563374645mtr.(c0 4 0.4160583942 0.4947122718S. 16 0.7449755474 0.4377280428Fyn 6 31.27007299-- 1147343713PIAIN 7 0.7299270073D-01 0.2610791095PHY Dm 2.343065693 0.5063995778
WI 1.186861314 0.7096827707SVP 10 2.411571832 0.3976152614'Put-ICC 11 0.71240876910-01 0.46008235440-01IlvFfICC 12 0.4319255474 0.9846640815D -01---'PF'Cr. 11 0,7868248179 0.1963778957nA/PW 14 7.171518952 14.38949504AW7mPF 15 0.1904223844 0.1584202399FrIcA 16 0.21408472950-01 0.11479490700-01ATP 4S 17 0.24441609E4D-01 0.17422144640-01sr.mt fl 0.5064407771 0.1071694530Erc16 19 6.204379562 4.462488557HNVAPF 23 0.41227493920-01 0.1063381657Fmnr, 21 0.7183501650 0.1871126820TMTNoc 27 0.7895912409 -,0.1492582915
361
'"0.^11'
%Atc 1
no,c,,Sqn 1.00000
3 4 5 6 7 8
-n.1116544-01 -0.117777 -0.8414690-01 -0.549443 0.235240 0.117308 0.771441 -1.278646 -0.251251 - 0.2.53636 0.158824 0.177/17 1.00000 -0.5340141-01 -0.286423 0.15/2480-01 0.128389 -0.8663030-01
1.00000 -0.116736 0.118669 -0.6607130-01 0.159821 1.00030 0.7240220-01 -0.2864800-01 -0.9927200-01
1.00000 -0.210368 -0.4264020-01 1.00000 0.8727980-01
11
V1.111C T.o,Cco
10 11 12 13 14 15
1.00000
16
1 -,.c71,c40 -n1 1.41,1491-4i 1.7)42550-01 0.172445 -0.263789 -0.3580980-01 0.3745700-01 -0.3079100-02 1111n .^114n,11
-1 .:10 711-11 -1.071,1,1-11
0.5195470-01 -1.111471
1,151 100 0.11'994
-0.2/7409 -0.7055/6D-02
0.6100190-01 -0.3419510 -01
0.9611240-01 -0.3750350-01
0.3132390-01 0.1564720-01
4 .1^7'111-1..17 -1.1?1..011-01 0.1762?70-01 -3.2117490-01 -0.2199160-01 0.115052 -0.3168800-01 S 1.100/450 -n -1.9757291-11 -0.4564310-01 -0.2150500 -02 0.156696 0.6110000-01
cp.177317 -0.7120020-01 -0.5094520 -015
_."4c01n-11 6,4609341 -01 0.101000 -0.124465 0.0798660-01 0.177848 0.8149260-02 0.2889390-01 -".144704n_11 -1.410.860)-01 -0.1143Q41-01 0.188130 -0.148184 -0.2140130-01 0.145039 -0.139114
4 ,071,4..1n-11 1.111110 0.491172 -0.117243 -0.595490 0.105344 0.274466 -0.272956 1.0m111 1.114254 1.514401 - 1.153294 -0.7279170-01 -0.297558 -0.348193 0.170842
1.00011 0.403151 0.245157 -'0.527995 -0.268350 -0.260920 -0.304105 1.00000 -0.272425 -0.684200 =0.8620500-01 -0.8219180-01 -0.4087370-01
1.00000 -0.342582 -0.5525270-01 -0.3124340-01 -0.255699 11 1.00000 0.8914000-01 -0.7260130-01 0.266827 !4 1.00000 0.509272 0.168116 IS 1.00000 -0.198695
1.00000
v4"t 1117 17 11 19 20 21 22
c
7
g
1^ 11 1'
14 IC
17
;
71
r.11-776 -1,4764^91-01 0.427105 '0.464170)-01 -0.146103 -0.211470 ..114<10 -1.1non,7 -n.4',456'0-11 0.194787 -0.211576 -0.249223
-.14211,01-12 1.50c.0470-11 -4.274499 1.605/61 0.5416420-01 0.3285570-01
P.11'4411'1-11 .1.170554 -0.,757/4 0.8717121-01 -0.4.142950-01 -0.3840190-01 .......1c4,..0 1
. I r-,1 11 0.817471 -1.176424 0.163124 0.178346
-^.117011 1,441e401-01 -0.?1,1605 -1.3547443-01 0.3209270-01 0.7105670-01 ^.17/144 -".Iti4i41-01 n.l"1704 1.302421 -0.5651970-01 -0.7696930-01 1%141,11 ..0091067 -0.7)1)100-01 1.157332 - 0.69119) -0.720173 -.141177 1.20o.470 -0.4141710 -01 -0.044510)-01 -0.157058 -0.3834080-01
^41.,csn-11 1.77S034 _0,109514 -3.145143 ,0.404078 -0.382289 1,,:577Jo -1.q17.470-11_-n.61,9070-01 -3.4851701-01 -0.862490 -0.772986 x.114547 1.111116 n:4401,.140-11 ".8601431 -01 0.156696 0.124999 .....51710 1.15e071 1.117,?400-01 -1.4116221-01 0.869236 0.878479
.^.1'101%cn-11 -1.44411A 0,01454n-01 .1.2521161-01 0.644/770-01 0.5425800-01 roinc49 -1.11x.717 _0.47.1170-11 0.407043 -0.9263780-01 -0.141468
_m5,*-120-11 -1.14^018 -n.2445410-01 -0.107282 0.145440 0.169777 1,014,s0 -^.264541 - 0.173617 3.211506 -0.602181 -0.584510
1.91900 ' 0.7422770-01' 0.3483584-01 0.253749 0.289834 1.00000 -1.104363 0.101165 0.105464
1.00000 -0.3991580-01 -0.6501400-01
. . 1.00000 0.987758
1.00000
1 Y"..
Zs
345
Occupation Stratum 1-3 Cross Race-Sex
VARI ABLE
SCHOOL 1
WAGE 2
UNION 3
MIG<53 4S 16 5
EXPRACESEXTRAIN /
MEAN
8.8050541522.266895337
0.50180505420.48014440430.411.552346630.91335743
0.267148014401.3285198556Os 79422382670-01
STANDA2D DEVIATION
2.9863657110.89103776410.50090171400.50050985400.4930355431
11.696946300.44327113650.47052504260.2708862774
PHY D,/ 10 3.158844765 0.7542412214NEG WI. 11 1.357430722 0.731331.2472SVP 12 2.251985560 0.5039388620BMOC: 13 0.1444620939 0.93681986960-01
04FOC: 14 0.2284296029 0.2387936304ZBFOC: 15 0.65581227440-01 0.4130958844DA/PW 16 19.81431531 37.11690734AVEMPF 17 3.3177830325 0.2611143641FDSL 18 0.32736540600-01 0.599998435870-01.ATPFAS 19 0.39776173290-01 0.17246831550-61%MININI 20 3.3590887916 0.1878449060EDS16 21 3.122743682 4.294345252UN,-;MPF 22 0.2012516652 0.2839610)08ZFMOC: 23 3.2940108333 0.3066052611tMINOC_ 24 0.4384,729242 0.2526728368 1'
RA *&EX 25 0.9386281598D -01 0.2921656195
363
CORRELATION MATRIX
VARIABLE
Numsal
1 2 3 4 5 6 7
1
1.10300
2
0.251772
1.30000
3
0.123764
0.392999
1.00000
4
- 3.7774350 -01
-1.9129260-.01
- 3.1069550 -01
1.30330
5
- 0.341516
..-0.378463
...0.223098
-3.113598
1.30003
S
- 0.522938
- 0.231373
-3.3893210.-01
0.3993220 -31
0.8348670 -01
1.33)31
7
- 0.132948
.0.243278
0.1413840...01
- 0.8666553-02
0.357206
3.1775740-01
1.00000
8
...0.6712210-33
- 0.483153
- 0.213359
0.4721010 -32
0.8656630 -01
0.144754
0.2934980-01
1.13333
VARIABLE
9
N4Matil.
10
11
12
13
15
16
13.278979
.-0.535448031
- 0.7581470 -01
3.6793950.-01
-3.112533
0.4255730-.01
-0.5:140473...31
-3.122453
23.194320
0.265983
- 0.9353510 -01
- 3.2108650 -01
3.1366.7
..-0.379372
- 3.411184
3.132669
3-.3.1e50390.02
0.141043
0.2295910-01
.-3.110896
0.2414940-01
-3.116945
- 0.176126
0.6527950-.31
..-3.6849670.01
0'.7556903-01
0.3430490-01
3.6628050...7.01
3.3628240 -01
- 3.1423250 -01
-.3.67:6543-.31
-0.2328853 -01
5-.3.134990
0.1843230 -01
0.7291910 -01
....3.8205180.-01
0.101616
3.3132330...31
0.3994700-01
-3.4031230-31
6-0.323284
....0.6660800.-01
....3.1796800.01
..-3.3285540....01
- 0.3798750 -01
0.3495680 -01
0.115627
0.5373683.-01
7 80.3387780-.01
- 3.6331790 -01
0.130194
- 0.484483
0.184998
..0.7921580...01
- 3.156471
3.143165
0,224333
- 3.451426
- 3.5657160 -01
0.617296
0.152236
0.472324
.
-3.1812380-.01
- 0.195158
91.03000
....0.26:0420..01
- 0.5235800 -01
3.136857
....0.8803510...03
3.3727530 -02
-0.6159100...01
- 3.3222820 -01
10
1.33000
0.507577
....3.632832
'0.735298
....0.671355
- 0.591319
3.243653
11
1.00000
-3.386817
0.492518
...0.231153
0.7799410 -01
2.659E370...01
12
1.00300
- 0.511932
0.216788
0.111965
- 3.228693
13
1.30000
- 0.693)7d
- 3.374247
3.179973
14
1.03)03
0.447923
t.-0.21801
'15
1.00000
-0.188547
1E.
.,.
1.03000
Cr,
vAA1.481.9
17
18
19
20
21
22
23
24
MumaEa
13.137572
0.601.53-01
0.217391
-0.5983190-01
0.4933610-01
3.116148
3.1456020-31 -0.24)4430-31
20.399659
0.137833
3.318402
-3.509472
-0.289846
0.441/73
-3.447294
- 0.503227
33.320646
0.884910-01
0.335484
-0.234721
-0.146645
0.707455
-0.156347
-0.18)401
40.5751480-31 -0.3752710-01 -0.6896590-02 -3.7141950-01
-3.152261
0.1333900 -01 -0.3596870-01 - 0.3319390-01
5-0.142854
0.5568710-01 -0.155356
3.7721130-01
0.07696
-0.193)81
0.3869690-31
0.6487483 -01
6-3.7778200-01 -0.7943490-01 -0.117378
-0.117372
-0.535;950-01
3.6989390-31
3.7171600-31
7-3.7668880-01 -0.8205453 -01 -3.151674
D.125203
3.5370380-01
3.277735
0.1271510 -01
0.4217430-02
0.8828130-31
8- 3.243260
-0.135043
-0.103418
3.521601
0.6575573-01 - 3.271759
3.654993
0.627427
93.138626
-0.7240730-01
0.3507240-01 -3.5998590-01 -3.2386923-01
0.8341840-01 -0.1981560-31 -0.2437150-31
10
0.304624
-0.1425273 -01
0.193480
-3.572653
-0.2505630-01
0.250554
-3.741145
-0.626719
11
-0.2937910-01 -0.101208
-0.5527970-01 -0.1903890-01
0.4251070-01
0.3173510-31 -3./51260
-0.9376120-33
12
- 3.159227
0.18)564
-0.7748710-01
0.228782
-0.5108423-01 -0.113561
0.210141
0.8518950-31
13
3.7119320 -01 -0.7944850-01 -0.2423800-01 -3.339414
3.2331950-01
0.9303710-01 -3.677837
-0.451755
14
-3.233384
-0.143196
-0.9144440-01
3.535278
0.i773660 -01 -0.232315
3.944054
0.883591
15
-3.323755
-0.8593160-01 -0.370892
3.619209
0.5299880-01 -0.253133
0.717721
0.732163
16
0.315218
-0.214538
17
1.03000
0.101143
0.526827
-3.448590
-0.8992020-01
0.773,355
-0.231791
-0.337894
0.1155500-01 -0.3449350-01 *.3.255320
-0.7338210-01
0.281233
1.00000
-3.394771
-0.5727480-01
0.451785
0.9328950-01
0.126107
:30.M42:72
-3.210980
-0.203253
--3.264999
19
1.'50000
-0.1330860-01 -3.233932
C.:
18
tin
20
1.00000
0.8403410-01 -0.398333
0.645295
0.657193
Cri
21
1.33003
-0.130543
3.8785793-0/
0.115257
22
-3.255045
1.00333
^0.274973
23
1.00000
.0.962133
24
1.00003
VAR1.481E
25
taim8E.1
t-0.6200400-01
2-1.301137
3-0.5057720-01
4-0.3676300-01
50.133308
60.7342203-01
70.533067
80.460135
90.4283500-01
10
-0.314533
11
0.4591230-01
.4--.4
12
0.129154
13
-0.216303
14
0.285380
15
0.474231
16
-3.124497
17
-3.207867
18
-0.8813860-01
19
-3.238292
20
0.394222
21
3.9474440-o1
22
-0.135682
..:3
0.397174
24
0.401751
25
1.03000
I
Cn 0
W .4".
.tr)
...4
1U
) 7J C, A (6 (..n
(1) X
348
Occupation Stratum 5 White Males
nal:ABLE MEAN STANDARD DEVIATION
SCHOOL 1 9.671171171 2.730147298WAGE 2 2.666146649 0.8895596554UN MEM 3 0,5990990991 C.4906336086
4 0.5112612613 0.5004370408_M1G<50S.AT16 5 0.2545045045 0.4360736756EXP 6 29.76603604 12.09207279TRAIN 7 0.1328828629 0.3398310226PHI DM 8 2.110360360 0.3478159601NEG WT 9 0.31531531530-01 0.1749463546GED 10 2.532882683 0.1126325025SyP 11 2.929729730 0.1680291250%6KOCC 12 0.7407432432D-01 0.4341615898D -01%41COC 13 0.M9346847 0.1113156607-A4irocc 14 0.2549549550D-01 0.6611944693D-02DA/PW 15 30.70714391 73.18803550NYL/PW 16 4.329228404 5.144872989AVIAN.. 17 0.4467765015 0.2752137339kDSL 18 0.3552091090D-01 0.4074778927D-01PDLEX 19 0.7086173176D-01 0.58541519980-01ATPFAS 20 0.4969414414D-01 0. 16 19400692D-0 1MININD 21 0.2845968977 0.1440114972EDS16- 22 2.317567568 4.251204310UNXMPF 23 0.3005274024 0.3264232519%FMOCC 24 0.3054301802 0.1134623060MINUCC 25 0,3765045045 0.9497692474D-01
:.366
00000'1 .
MIS/Sr° 00000"i1.0-0SE995b'0 o-aLELLWo 00000't /10-079EMS'0 tO-CtC4r09'0 E6C06Z'O 00000'110-0E9SZOE'0- ZO-aLLLLO9'0- 9999/.1'0 eflfl09'0- 00000'1
10-aZSLfl92'0- 1.9-069ZSICO t9COSS.0 9Ltal'O- SSZEWO 00000'1
10-ata9t6"0- tO-G060627'0- OESOtf'0- SLCM1'0- ZE67,SF.0 20-CL,0Lf/S'0 00000'1
1.0-016119Z4'0- to-aeitaSE'o- EL116S'O- 116L017"0- 10-09flUZL'0- 6E069E'0- r0-0,57.)69'0- CCOCO't .
to-acinffuo- t0-092ZS0L*0- 9901/.1'0- nssta'o- feOrtS0 10-asin58170- 9SE'0 10-09LLP01,'0tO-assu.SEo- t0-c695c0Vo- t0-ORR699Z'O 66ELL1'0 oh141Z1"0- to-nrr*opt'o- tp-n*Do65vo- i0-c0too2r'n-
6SE601'0- 1.0-096LLW0- 1.0-a0OUSC'0- FC9161.0- no9cPL'0 i0-(165cEWn opftwo 1C-CtP6eLf'0
10-a8tArt6N) 107(710059L'0 10-4716069010 to-0416000- t0-000q49r'r, 10-4970Lbs'0 10-00ILLsc'41- fo-e06f,70b,0-
10-06Z799C0- 10-COthS6t'0- 10-0Uptt7'0 70-096EOSS'C 10-0909t0t*0- 10 -O 04(060'0 70-att4tS7'0- le-cf7ros'o-
tna81'0 10-0thL8LA'0 70$111.0 1.0-06(0077'1., 10-01671.07'0 t0-nrtr7wo 11171-1 '0- 1 0-0 tzt Pei A "o-91(6SUO 069661"0 1.0-7,9118ZOS'0- 7t6Sit'0 696ft91'0- 10-rrrtc016'n- fLrt,CL'f.- tc-rcfC01"0-
hattSt'0 ltchOl'0 Z(riSt.o- EtoLWO 977FP7-0- atttf0.t'0- 777f41'0- tf-Gef7;1P'0
91 St 171
00000'1.Z0-0tWEE6'0 00000'110-0StL999'0 19LCZE.0- 00000'110-0E95915'0- t0-00SZZ0E°0 10-0919LSV010-050179E1'0- 10-eZ6U9V0- 10-0az9L19'0to-anozwo 10-0911(S17.0- to-azoSorC"o
199Z01'0- L6ftSWO fzEbt141'0-165811'0- 91,Z0LZ'0 6E0/69'0-
9 L 9
£1 71 11 Ot
00000'I811ZW0- 00000'1t16rrl'o-0-0806SECo 00300%7607P1'0- 961S0t'C- Ofn.fivel9h011t'0- t0-36S0919'0- 70-41996S1t'0
S
00030'a(91)9C0
i2r1ZP9
71e%Irl'A
1
0
Se.
C
Oconotrt
1.'rSalwlior:ryA
40T4Y1=4m0:1
YnaLA0Li..4811ci1
1
2
11
3.1146010.3J35460...1314
10
0.1117590.1):46,0.141441
19
0.17'26028.2119010.125909
4 -0.6570530-31 -0.4341)30-01 -0.4152540-01i 3.11/7)00-81 -0.311720-01 -0.1/55100-01o -0.9084/90-31 -8.1144.5 -0.1076157 0.411/508-3 2 0.3774200-01 -0.0260070-038 -J.45441l8 -J1 c.76)1(5u-01 3.8372)10-01V -0.322/14.3-31 -C.24/2030-01 -0.8'267900-02
10 -3.9069610-01 -0.111358 -0.112297II -0.119561 -0.5317630-01 -0.126176
12 -3.0234790-81 -0.111680 -0.15470313 ,0.1737664 C.100294 0.14240814 -0.1623190-31 -0.21.4907D-01 -0.5174150-0111 0.3uio36 O. 117125 0.1115060
I. 0.570244 0.240794 0.2414671713
1.00000 0.2342571.00000
0.2000350.8)0208
1* 1.00000
1021222324
i0
10'1=3 0.230514
0.514272D-01-0.1104470-01-0.162510-0.2791490-010.2311920-01-0.610052D-01-0.272362-0.5098100-01-0.09/6130-010.112076 .
0.7167100-01-0.11.03300.2477930.3599710.2156180.3176121.00000
VAidArLE 254034:65
1 J.0749850-812 3.1181)43 3.4,7679J-014 0.1347100-82S 0.s/1901J-42o -1.1421047 0. 15413J3 3.5,36840-31
-8.5,25101J -0.2v3470-C111 -0.4321058-J112 -.0.23364413 2.911101514 J.4»73815 3.02.4158-J210 3.4314328-J11/. 3.1+001/
1) 0.4430153-3128 3.4:343660-0121 -0.1272300-0122 3.2527.6D-0123 0.590003D-8124 3.92339625 1.00000
11
-0.153175-0./54646-0.2201100.4116100-010.7219230-01.0.61237130-01
-0.2657510-01-0.4307160-010.3427710-010.3540170.701203D-010.3726)20-01
-0.2133440-01-0.6626080-01-0.202394
%:°0:.'i31441Z491
-0.252498-0.223267-0.361329
1.00030
22
0.8545230-01-0.7401100-01-0.9701420-01-0.1412130.934362
-0.8335160-010.7541200-01-0.7710780-01-0.3777520-01-0.5493650-010.2404230-01
-0.2584420-010.2052150-010.5261210-010.1363420.1234970.644569D-01
-0.1976670-010.1608100-010.3123930 -020.4996493...011.00300
4V
23
0.6311910-010.2264120.7493920.6436640-03-0.2645800-01-0.1953470-01-0.3747643 -010.416955D-01-0.9481524 -01-0.120049-0.1907540-010.2188420-010.3420850-010.7327390-610.2809880.4648960.7059760.2803600.257986.0.367930- 0.3216080.9074440-02.1,00000
24
0.1644710.1696830.260836D-010.6692770-02-0.109839D-02-0.1900400.175727
-0.153758-0.434643-0.160028-0.102475-0.5824120.9984520.3517280.8314613-020.5271930-010.1537550.9713220-.010.136774
- 0.2490869 -010.3104780 -01-0.175560-011.00000
O00CIDCr
0
cr)
rt-t
rtC
%.J1
ti
rt(3
u
0rr
351
Occupation Stratum 5 Black Males
VARIABLE
SCHUOL 1
WAGE 2
UN MEM 3
MIG<50 4
S.AT16 5
EXP -6TRAIN 7
PHY OM 8
NEG WT 9
GED 10
SVP 11
MEAN_
- 8.3393501812.385234657
0.5523465704( 0.400722021.7
0.718411552330.02527076"0.12274368232.19/i9458480.14440433210-012.5516245492.946931408
STl iDARo OEVIATIOV_
_3.485009868y.80137090364981522915
0.49093173610.450587784712.65896359
0.3287366838"0.42337893820.11951348030.12411671920.1306324041
8M0CC 12 0.897833935003 -01 0.62065621060-01--7,WFOCC 13 0.2651732852 0.1161655441%3F000 14 0.26772563180-01 0.63970870520-02DA/PW 15 22.20990909 39.13718419NYL/PW 16 3.600650095 3.657186912AVEMPF 17 0.4236900120 0.2953215854FDSL 18 0.38481035710-01 0.68271206250-01"
FDSLEX 19 0.61,083640790-01 0.78777844050-01ATPFAS 20 0.45847653430-01 0.18336128010-01MINING 21 0.3057716861 0.1555410360EDS16 22 - 5.501805054 4.538582225UNXMPF 23 0.2807524669 0.3346285613%FMCCC 24 0.2919458484 0.1191812678MINUCC 25 0.3817292419 0.91073721790-01
369
C044ELATICN MATRIX
vAKtAbLE
NUMBLK
12
34
67
8
q1.00000
0.264135
0.139996
-0.103064
-0.312712
-0.565397
0.232328
-0.106389
21.00000
0.570692
-0.174990
-0.276255
-0.5536530-01
0.162318
-0.3099750-01
'
3_
,..
--
1.00000-
-0.108306 --
-0.240785
-0.4588760-01
0.115497
-0.4856030-01
41.00000
-0.126835
0.380089D-01
-0.81.7211
DI.-0.4600220-01
51.00000
0.161324
-0.132722
',..0.6061357U-01,
61
1.00000
-0:262814
0.175520
.7
1.00000
-0.1635250-01
,
8t4
1.00000
.41
NI
4:)
VAR150LE
tiumaHk
I 2 3 4.--
5'
b 7 8 910
121314
15 16
9
-0.2050710-01
-4.806t420-81
-0.134457
0.8627480-01
0.8501250-02
0.2610110-01
-0.4527780 -01
-0.5583660-01
1.00000
.-- ------.
10
-0.104309
-0.207776
-0.257756
0.1603540-01
0.1730750-02
0.175116
- 3.102585
0.552443
- 0.5043820 -01
1.00000
'---
11
-0.4943520-01
-0.48t2590-02
-o.igesoup-ca
0.3336740-01
0.2219970-01
-0.6601153-01
0.7629920-G1
-0.316697
0.414881
-0.246062
1.00000
".0.275046"-*'-`
12
;0.215242
-0.188413
-0.191349
0.670704J-01
0.9653340-01
0.229475
-0.6706013-01
0.710469
0.757755001
0.679542
1.00000
13
0.233809
0.2104 f1
0.1884'O
-0.140802
0.9034390-02
-0.t44747
0.4915710-01
-3.299122
0.249672
-0.363919
0./49786 - --
- 0.679247
1.00000
14
-0.1619u30-01
0.9562380-02
-0.8188600-02
-0.7792590-02
0.158707
0.2425840-01
0.2538330-01
-0.184235
- 0.450638
-0.105630
0.402164
0.4116440-01
0.450006
1.00000
15
-0.111416D
0.I62213-01
-0.126132
,
0.168385] 01
0.112849
0.4015351 02
-0.2252161 01
-0.69606st DI
0.2757731)"1
-0.115032
0.358716(..
.
-0.503984f-01
0.255081t.02
0.15519 1.-01
1.00000'
.
16
0.100481
0.334962
0.149023
-0.2048750-01
-0.104782
-0.6757560-01
0.5438470-01
0.3707150-01
-0.2816950-01
-0.20042
0.453568001
-3.147826
0.133718
-0.5687620-01
0.645331
1.00000
°
VA
8111
LE11
N,U
N8E
R
0.1J58-)
20.
487.
3,11
30.118745
0.95757,0 -02
5-0.7316390-31
60.4270030-01
70.132335
0.1',1544
9C.6L.2210-31
10
-0 195163
11
-0.9506010-01
12
-0.113676
13
9.9856930-31
14
;A19750450-31
15
(0.343411
16
0.696181
17
L.00000
18
1920
212Z23
24
114
0.5037510 -01
3:1/1617
.J.1210,85
0.2144200 -31
-0.121166
-0.5066420-02
0.305422J-31
0.2466040-01
-0.1553083-01
-3.401357u-01
0.5629300-C
- 0.109699
0.75542507.01
-0,128244 it
-0.341090042
0.165074
0.226,17
'
1.00000
et
19
20
21
0.9079440-01
0.22055
0.1649990-01
0.260958
0 43,-(22
-0.316254
0.149'54
0.6n4.58
-0.230515
0.1,1'9100-01 -0.1622860-01
0.6240110-31
-0.151844*
-0.265481
-0.353-..U-01
-0.4533740.01 -0 446843-01,-0.6117920-01
0.7067403-01
0.6394320-01 -0.9553490-02
0.10332,0-01
0.399168301
0.448.460-01
-0.2'9187W-01 -0.1035210-01
0.9581340-01
-0.8136763-01 -0.306390
0.345780
0.65449.0-02 -0..1223/1
-0.1320370 -01
-0.144810
-0:160854
0.130327
0.105!5:
0.111593
- 0.5573510 -01
- 0.1352.
-0.15:,842
-0.7387710-03
-0.164124.-01.-0.2343630-01 -0.200744
0.209108
0.576451
-0.335925
0.285080
0.542072
70.359535
0.963522
0.6948573-01 -0.272305
1.00000
0.172672
-0.291672
1.00000
1.00000
".
22
0.311124
-0.104256
-0.132652
-0.181,011
0.160311
-,.197103
-0.1714790-01
0.1101520T01
0.26.104J-0.
-0.3%93360-01
0.1010020-02
-0,4505530-01
0.151304
0.113139
0.2277260 -01
-0.6282330-01
-0.1215400-01
-0.104007
-0.117966
-0.13
87
0.26 5 70-02
1.(10000
VARIALSLE
h0303t8
.
2 3 4 5 6 7 a 910
11 12
1314 15
16171419
21
2223
2425
25
0.1504
0.14,155
U,10',444
-0.134434
0.8845750-01
0.2625740-01
0.1876279-01
0.8970150-01
-0.401752
-C.8503050-02
0.3186110-01
-C.182007
0.844222
0.672264
-C.3650560-01
0.654210D-01
0.4,91310-01
0.1258053-01
0.2670100-01
0.880960D-02
C.1767330-01
C.170232
0.130691
0.858945
1.00000
23
C.990766001
0..2..7324
0.5;0193
0.205,,49'
0.756678
0.163141
-0.530,2220.'01 -0.117667
-0.199149
'0.1732440-C1
0.612111001 -0.139564
-- .129609
0.4927540-01
4994520-01.-0.301442
-0.101:41
-0.267542
-0.20184
-0.360381
-0.5639000-01
0.167562
:0.174218
.-0.659850
0.198396
'0.998851
-0.5174040-01
0.492295
-0.167.2280-01 -0.1653260-02
0.447120
,0,127281
,
0.725547
0.9122710 =01
0.227042
0.6674750 -01
0.300084
0.500346
-0.338649
-0.1368
1.00000"--(1-
0.9581620-01
0.100915
-0.5436440-01
0,15354e
0.190596
1.00000
354
Occupation Stratum 5 White Females
VARIABLE
SCHOOLWAGEUN'MEMMIG<50S.AT16EXPTRAIN
1
2
3
4
5
6
7
PHY DM 8
NEG WT ` 9
GEO 10
SVP 11
%8MOCC 12
%WFOCC 13
V3FOCC 14
OA/PW 15
NYL/P4 16
AVEMPF 17
FOSL 18
FUSLEX 19
ATPFAS 20-MINING 21
E0S16 22
UNXMPF 23
7,FMOCC 24MINCCC 2.5
MEAN
9.9451627122.0097288140.41355932200.50847457630.274576271229.93220339
0.14576271190-012.010169492o.a2.604'4C67802.9203389830.46284745760-01U.45688135590.28936440680-0115.414216C42:1541274230;39423073450.29577245110-010.75153344610-010.44581355S30-010.45075372582.742372881
0.18641340510.48586779660.5321525424
372
STANDA/R0 DEVIATION
2.2731291950.-42205766610.49330815360.50077761260.447058877110.35141780
0.26315279720.51826992260.00.18739415320.18329900210.27311597660-01f'.24214270860.10933619400-0119.785093172.721771670
J.z336196155,0.33049555440-010.:5433353090010.1697)918860-010,1129861367 \4.639530068
0,29186721780.24796618870.22 3212915
355
Occupation Stratum 5 White Females
.4 .4 .4 .4 .- .- - .4 44 .40 0 C) 0 CD 0 C) C) 0 0I I I
1 ' 1
I I I
C5 C) C) 0 0 C) CI C) 0,, o, -400 4, r. 0 44. 07 44 .4 C) 77 4.. 44. = ,r 0 r- A el 0u " -. 4. 4. eft .4 P . c , N7 I) ... .1, C. O. C. ID .... c. NI Cr 0 0
CO 0. , ., .4 .4 c) 1.4 0 4c. 'or .4 ..7. 4. 44 0 0 .7 -1 uN r. v) tO .73 0.09O,00000 . 5 5 TN ..r no a 0 - c . - 4 4 . o 3 ,.., .0 4: 0.... f... 4. 0 t 1 0 .0 .".". 0 -.J 13 0 r- C) .r. ,..) 0 0 ..., e4 .4 c. 044 -4 .4 0. 444.44.4 44 J? .... :V ...4... ..40,3 .... . 1.... r4 .0. . . . . 1. .., t0000000000000o .40 C3 0 0 0 CI 0 0II III C7 C) C) 0 0 0 CD 0 0 C) 0 C) CD4 9)
I I I-
.4 ..I N 4 ..I ... 4 .. .40.000 0 00 0'11:1 I - 4 41- . .).
C5 0 .".. 0 0 00 0P,,4 13 43 "si 4 '0 rm N 7 .0 4.1 0 1.. rg -. 4, N .. PO,4 rt., A ... 1) fs 0 0 0 4. "s.1 4.,.% .4 4. 33 4 7 J 0 4 0t. r ....NN..! r 0 V* r s ON 7 - tO .0 0 rl .4 ... 0 .4 0-7-7.0-.0.00 .4 N. N. NT 4. 4. .4 oft P./ - P./ P.1 A 07 A " 1 A A 0 0 .0 ,,J .0 4 tO 7 ... .r ,-, str v 0 NO- 2 . ^ 4 .4 1 4 1 4 - . . n J1 .4 CO .-. ^. ..r. 0 ..r cr - ,r eu o
...Ir.. -t 00000 900. . . . 0 0 .40 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0 0
I I I 1 1
.4 .4 .4 .4 ..... ...000 0 0 0 0I I I I I I I0 0 0 CD C) C) C)
01 4.10e.a3 0.4 .1 A ^,- 07 CD 1 N1 '2 7. '4741 S 2Z t; MO ':7. 43." T. 8'
,45 .-.. 0 J1, r00 ..). -4 J'0 .... 41 r. P. .4 ,"4 (.9 4' .t. 044 c) 4 CD .4. CI .4 41 ,r 4) 4. ... eu .1 ,,,I .. X) Cr 43 C)
V. ... e. . ) 44. 3 PAv n. O. r4 A 0 0 ,4 ?. 4, .0 7 .0 .- 0.... .4 .-.. Nt .4 A.. NO 44. 4 Cr tin
O 0.00 w4 o0ool000000loae w40 0 0 0 0 0 0 0 C._ 0 0 0 0 0 0 0 0 0
1 1 I i I ; I I
.4 . -. ru . .4
I I0 0 0 00 0 01111 1
0 -7.N 40f.?0
N2 2 R un 40 .4 44 4, C)N. CP 44)0 .........,^1..... !..4%.11 1... .70%0
IA In A7. 4.0 0 ,.., .., p.. un t. .1r1 V. O AOPIP N P. 0 ... , .. 077 CO -7 en to 4 77 00..., .4 +J a. 4. 44 r .0 4 C) 0 1)- A .4 C)-4 e4 Cr N 0
441 .4 CP .7 .0 r- Pa un 0 41 Ma,. . 0 . 0.4 099000000000 "4
0 0 0 0I I
0 0 0 0 0 0 0 0 0 0 0 0I I I , I I
N .4 .014 ,...* N . 4 ...4 .4
I I0000III)000
4. t...c 7ii :.J. ii
.., 0 ,r (, ,D 0 ... r cD .4 C)4) .4 .0 ..4 4. f... 7 Cr CP CP CI
en CI a ,1CI 0 CD
4, O'S. C) . 4.4
4. 44 .. ). ..C, .0 ,.4 1: 7 ..T. 00 " ., .1. T. 1` I, '1 rT J0
c). Pa in.n C5 -7 r4 4.4 e.. 0 ..i.' ..., CD., ... a% CD ON 0 7 4 7 ' .., 0 0 4.0 0 0 0 0 0.400000 0 00300000.000
I I I I I I
0 orlf0j NC) 00000000I I I I I I II IC) 0 0 C) 0 C) C) 0 00 0 0 a% A ,3 N un 44 en an CP 0CP .00 ..5 0 .4 c. I.- 45. )44. 4) un 0
R1 un NJ 0) rm. ,. , . N .4. c) 4) Cr C)..4 +0 . 7 f+ 0P .4 -0 r. an NOpg u1 0 44 44 4. .0 44 ra 'V 4: J00. Pa 44 I) ^4 ") N! J 4..i V, C) .44 .4 . . . . 0 4 -
C 0 0 0 0 0 0 0 0 0 0 0I I I / I I
.. 4- .-C) 0 C)I I I
C5 0 C3.1 0 v. 0 IN, .,, .. .4 .. 4) 0770 - O0 ,N .7 ,^1 P... 0
N ^'0 0 AN .11 ..) 41 ..! NT - V 0-4 CJ -. N N 4 0 ..) ../. !.. 77 0P 0 ,p ,,, zo oaniOUN 0
, - - < si ." ON CO.4 9 99 904
O 0 0 0 0 0 0 0 0 0i I I I I
... 0CC CD.. ..., Cf.
4 8X CD 0 0 0 0 0 0 0 0 0X 4 e
LI kJ C) 0 0 C) 0 0 0 0
.4 w4 ...k Cc 4 ,Y .4 A ..cs 455kutat 4 0 .4 IV el 4, Vs A P. 0 .1 na .... r.. " 4 N.0 P- 35 7 0 4 NI "I 4 uN AX - 1" - XCt eX 0 a 9O 4 Z 4X-/ 5. >
373
',439
14.a
lEh6
46t.t
t....
5 7 a 9
10
it12
13 15 te 17
'
1 19
20
212223 24tr
.LE
17
0.2391:72
0.401646
:.1616/7
0.1199460-31
-0.d227tIo-J2
-0.4111140-01
0.126699
-0.136326
6.0
u.s970.-3
C.66,1210-31
-0.230499
6.191091
-0.277639
0.705330
0.72
1180
1.00000
25
Id
-3.2311493-01
0.160513
0.1023.,4
-3.120251
-J.5648,3,5'5-01
-0.122...00-01
-J.7:3240 -01
0.146412
0.0
-J.294203
--.104144
J.119912
-0.242919
-0.134130
-0.251553
0.500987J-01
0.9014500-01
1.COLUO
19
-0.1710630-01
(s.t69671
0.8819430-01
-0.151561
-0.113631
-C..7502140-01
-0.6902220-01
0.312674
0.0
-0.4)3721
-0.113359
0.164376
-0.331169
-0.22129/
-0.270926
0.4950110-01
0.206.181
-0.812021
1.00000
20
0.106163
0.406870
0.199369
0.216/41) -01
-0.112642
-3.9186290-01
-0.311325J-02
0.140014
0.0
-0.275640
-0.102277
0.65'040-01
-0.246665
-0.331743
0.105538
0.639349
0.549196
.-351074
0.506030
1.00000
-0.1311600-01
-0.331060
,
-0.146241
0.8101790-01
0.154020
0-8413800 -01
J.4356240-01
-0.219991
0.3
0.2/2286
0.8550540-01
-0.4580220-01
0.201660
J.189869
-0.862/890-01
-U.394302
-0.416014
-0.321063
- 8.401020
-0.658W37.
1.00000
22
0.161542
.0.103524
-0.161350
0.4486220-01
0.962397
-0.131664
0.8824440-01
-0.9192620-01
0.0
0.1069960-01
-0.6211200-01
-0.1123220-01
-0.1169590-01
-0.9540480-01
0.9284760-01
0.1114760-01
0.1226160 -02
-0.35,7790-01
-0.111934
-0.108613
0.151716
1.00000
23
0.35067?0-01
0.35°742
0.761075
-0.26.16040-01
-0.905530-61
0.1324330-01
0.8414110-01
0.3114110-01
0.0
0.1569400-02
0.1526320-01
-0.9C41360-01
0.3770370-01
-0.136292
'0.351011
0.39/502
0.569496
0.121411
0.135169
0.372361
-0.286496
-0.7570820-01
1.00000
24
J.341119
3.13499A
0.9740160-01
-J.7235950-02
-0.6490310-01
0.7694750-01
0.191012
0.606641
0.0
0.874561
0.6252560-01
0.811458
0.999287
0.548414
0.362,683
4.100226
0.6108990-01
J.243128
0.333149-
0.254543
0.21,1.155
0.1563150-01
0.26i9640-01
1.00000
,, C lc:
C, rs 0 D
itn re
.n W rt C a %..n
10.
1 s
3141
20.1 43541
3-3.9 665650-31
- 0.674°^i40-02
S-0.6603630-01
6-0
,724
4840
-01
10.1da!i4
8-0.513115
0.3
10
0.878,04
11
C.11C121
12,
-L.7oat:3
13
6.094adt
14
0.5691,8
15
0.376912
160.4256 510-01
G.:64519
td
-C.1418,0o
t-C
.145
11 /
2J
-C.2'0,64
0.225old
22
-0.1a4a49:i-01
23
0.1790620-31
24
0.997449
25
1.00030
'
=1-
rt 0 tD a W to 0 rr
357
Occupation Stratum 5 Black Females
VARIA6LE
SCHOOL 1
WAGE 2
UN MEM 3
MIG<50 4,
S.AT16 tEXP 6
MEAN
10.132911391.863670886
0.46d35443040.39240506330.639240506325.34177215
STANDARD DEVIATION
2.5012211800.67107229950.50058419470.48963:69750.481747805611.21462430
TRAIN 7 0.20506328880 12 0.37869999600 12
PHY OM 2.0-56962025' 0.-4259611780
NEG WT 9 0.0 0.0
GED 10 2.581645570 0.1599516677
SVP 11 2.956329114 0.1386916749
6MOCC 12 0.57063291140-01 0.33045940530-01
(XWFOCC .3 0.4154746835 0.2300809522
2:8H1cc 0.10556962030-01 0.102013959270-01
DA /Pw 15 10.09615030 15.44776192
NYL/PW 16 1.925543429 2.300089160
AVEMPF 17 0.3171822765 0.2863453859
FOSL 18 0.31807681290-01 0891578830-01FOSLEX 19 0.67439290550--01- 0.67260813240-01
ATPFAS 20 0.41330379750-01 0.15993094460-01
MINING 21 0.4802458420 0.2010955900
EDS16 22 6.291139241 5.171336726
UNXMPF 23 .0.1942523207 0.2957062699
tFMOCC 24 0.4460316456 0.2366932389
MINWC 25 0.5030949367 0.2136431534
J,
358
Occupation Stratum 5 Black Females
rj . ea .4 -4 N -I ++ .4 .0000 C C 00 00 CIiii I
cloc)c) c) ,I -. oO LI ,S
'",.. ;?, 8 ;4 :1; :4' 4 gc, .3° -. ... 0 .0 TA .0 el eV Ur 0 CP .-4 C)J _ - , . _ -
- . N0 h. 4 -. ' 0 1 0 - 1 c0 C) 4. r4 43 C).0
...r. t.24 .17 1 .4'3 r...-5 . C
43 41 .4 41 MOW 04 0O .4 ra .3. 0104 N+ ... .4 .4 .4 1... 0 4. .4 v. PM el 1.
0 .4 v.
0 0 C./ C) 0 C) 0 0 0 C) C) 0 (3 0 0 C) C) C) C) C) C) 01111 1 1 I 1 I i
..4 .4 N.. .. .4 .0 ,JO op 0 0 0 C) C) C)Mt 4 I I
q.10 t'. -.:340.4.4440 TIN P. ,4 " C) 0 C) 0 (IN 1 w. el 4 r. C) 4 re) CP n. C)NN c0 P.- CP 0 0
P. ,a ,.. ea .4 .4 el 4) a.N
ra 40 44 4. UN 0 4' CP 0 r4 ..1. r4 47 0.4 -1 0 ..... r4 .4 4) 0 uN a) 4N ru N0
A i.7) r3.40 5 W.- 8 "'u1r... 40 c0 -4 .)4 0 it." .41 I.- 0ra 0 r. .2 0 4T a3 0 4. 4N .4 CO C7.4 .4 .4 .4 r4 .-T P.T. el C) Jr.. r4 rn r4
1! -: rl rl 1 -: .: . . . . -c, c, c, C) C) CP1111 C) 0 C) C) C) 0 C) C3 C) C) 0 C) C) C2
.4 .1 .4 -+
1 i I i i 1 1 1
.4 .-4 0.0 0 0 (7 0 0 c)t t 1 1 I I I
j!:
0 0 00 ..1 4 ,o .7 ri n4 0 CP C) n4 C) 0W 1 . 4 . .4 40 C) 0 C ) . 4 " .0P 43 0 0 UN P. 4.. eV eV C)
a) 4. 43 0 N ... C3 1/ .3 .0 4P N CP .1 0rd% ..4 CO U, ,1 F4 44 ra 0 C) al UN 0.4 ea P. /3 4. C) .40 .4. al ra en 0 4) N uN 4. ... MO," ro, NI eN .-. .. 4., fq 0 .0 ,n r4 .0.4 - 0 0 14 .40 0 0 0 0 . loot ... e .4C) 4) 0 0 0 C) C) C) C) C) C) 0 0
1 1 I IIII I I I
..I .4 4 . NI0 0 0 0I I I 1
C)4 ..t r4 rn 0
0 0 . C)4 ,.. 10 1% ... 0, a) CO 44 0 ul C)
PIN 03 P- C)NJ 0. 0 c ,
. 3 ° .1 7 . rN C) . 3 0 4 4 rn 4 .0 0 C)41
%Ft CO O r-ni
0O -.., 2. .4, .4' .74 ll, c..;
OPg
. - 0 40.4
D C) C 0 ., J'0 4N CP 4. ...r c,... 4 I rl .. . .4
e4 e0 . 0 43 r4 r.
C) C) C) 0 0 0. 1....1 ;,..
0 C) 0 0 0 0 43 C) C) 0 C) C)11( 1 111 111 1 1
v.
I ?0 0 0I I
0 0 ,.... 0 CIuN .0 r4 0 .0 0. ,N1 IN (NJ e- 03 0 04. el JO ..'1 44,7 a) 4J 0 .... 0. 0.4. e- c3 0 eV , .7 3'r .+I rn a3 P.- t- 00 0 .0 0 4 * Z. " ti P. C 0 N CO .0 .0 00 .... Ws 0 ,.. .4' ... '... 0 0. 0 CO MM 0v. at . . .+0 rg .4 .4 .- ra r4 0 .4. e,,, .
. .4 . .0. ....off .4C) 0 0 0 0 0 0 C) 0 0 C) C) C) C)
1 1 1 1 I 1
.... - ... - ._. .. 0.,
o .0 0 C) 0 C) C)I III fi
0 0 030 00. 0 0 . aD .. .r ,r, .1 .1 - NO4 MO NJ 0 r- -4 0 41 4.4 OuN rn 0 ., .4 Cr C) .4 C) Cr ).- V, uN 00.00 .4 4 -. 0 4 C) 4. ,1 0 r C)
e. e) e... 0 ... as 0 ,..1 fsj .4. NOP. r1 03 -4 4 .. U% ... NO P. 0 .4
.. 1..40 0 0 0 0 0 0 C) 0 0 0 0
1 i 1 1 1
0 0 C 0.. g1 A,0 00... 0 J,.... .0 P.- r1 .3° CO .- 0
MU .. 7 4-. 4 4 0,4 e4 0N 00 C , c , 4, 4. r. ...r 0 1.0 (0 40
0 0 1'1 0 -+ r. n) 4 0 '0M0 0 4 0 r4 ,r e, D co 0rN *3 .... .... 0 Nn ,r0
a. e ..C) C) C) 0 0 0 0 0 el C7111)111
)cO
QC 004 0
. 0 0 0 <3 0 C) C3 C3 C)z .4 . . . . ,
1.1 (7 0 C) 0 0 0 0 0 43
P. (4.. 04.1 .1 CC - `C... 0 .0 1...../UJ at 4 .4 "4 0% ...t. 4 4) P. 0 41 . .. ea r1 .1. us .0 .... c01 7 47 .4 eV rN 4. r. 41m ..... z - x -. -..4.4.44...+X X 7 X0 0C z C .0...) > >
ys4f4,11EhuKatA
17 id 19 20 21 22 23 24
I 4..1 T',>,,,,0 0.',h)1110-01 3.145601 0.111313 -0.175940 0.181068 0.154715 0.1517752 0.4,i. '3 0.1S3418 0.216115 0.1'6128 -0.1,5316 0.1508510-02 0.470E141 , -0.1197733 0.11:1,o) 0.45777 , 01 0.110313 0.176717 -0.102438 -0.1610110-01 0.702115 -0.1732744 -C.4c:,%60-0. -0.4411010-01 -0."..41400-01 0.6102440-01 -0.2637290-01 -0.154666 -0.408371,1 01 -0.9091660-015 -0.14,1 )1 3.19',,JZ4J-01 0.p51o340 -02 -0.1234133 0.918i040 -03 0.916816 -0.0201610 Oi. 0.1666960 -016 -0.013K120-01 3.7141b0001 -6.1281270-01 0.'668173-12 -'0.1161570-01 -0.161185 0.1165600.01 -0.2464457 -0.113464 -0.191)289-01 -0.153/114 -01 -0.1119100-01 0,116982 -0.4138810-01 -0.4814420-01 -0.3531030-01& C.51,4810-01 -0.1279280-01 0.148154 0.104866 -0.299618 0.174581 -0.4079810-01 -0.15112074 0.' 0.0 0.0 0.0 0.0 0.0 0.0 0.010 C.8'16090-01 -0.1671560-01 -0.196121 -0.316911 0.170767 -0.6819170-01 0.331,5760 -01 0.09832111 -4.1721164,-01 -) 1431890-01 -0.1771700-01 -0.152060 0.160559 -0.8288360-01 0.2161480-01 0.1212570-0112 -C.704/440-01 1.150;320-:.! 0.4215810-01 0.128112 -0.5750409-01 -0.130783 -0.129114 -0.73148011 0.3998600 -01 -0.718160)-0) -0.182192 -0.141419 0.151272 0.3361920-01 0.1439910-01 0.9,044514 - 0.369259 -0.953/310-02 -3.140447 -0.53411 0.308393 0.111115 -0.180503 , 0.66066915 C.671823 -0.142104 -0.108416 0.100806 -0.23256) -0.126938 0.496814
s0.301748
16 C.694149 -0.6217600-01 0.108192 0.713234 -0.411100 -0.194919 0.437177 0.1645870-0117 1.30000 0.6224970-01 0.130663 0.592817 -0.558036 -0.7800480-01 0.749664 0.2194170-0118 1.00000 0.623152 -0.1609580-01 -0.141136 0.8738230-01 0.1623800-01 -0.701767.3 11Iv 1.1 .,100 0.300802 -- 3416.5 0.8463820 -01 0.14a88. -0.18115620 1.00000 -0.657822 -0.7114460-01 0.391103 -0.15505321 1.00000 -0.3933130-01 -0,347218 ' 0.16617622 1.00000 -0.3897450-01 0.3790570-0123 1.00000 0.2133570 -03.24
f 1.00000
121.1 14131f 25humt10.
C.14 ":e."2 -0.1427513 -0.16014;4 -C.821c0-015 C.348514i>-02
7
8 -0.152118
10 C.,:771211 0,53241,0-0112 -:.457.'111ik 0.64t,Is C.2./71,6lb 47.4745,.C3 -03
17 C13
19 -Z..1'4616120 -0.114418
0.1,75114
22 C-21766%,-0123 4.3474100-0224 C.994456
25 1.00000
C)
0C't
360
Occupation Stratum 5 Cross Race-Sex
VARIABLE 11EAll STANDARD DEVIATION
SCHOOL 1 9.687727825 2.627930192WAGE 2 2.480206561 0.9147806731UNION 3 0.5212636695' 0.4998514205MIG<50 4 0.5030376671 0.5002948107S 16 5 0.3037667072 0.4E01627809EXP 6 29.69380316 11.61176717RACE 7 0.1117861482 0.3152947765SEX 8 0,3888213852 0.4877790632TRAIN" 9 0.1081409478 0.3107471877PHY DM 10 2.077764277 0.4229186710NEG WT 11 0.1701093560D-01 0.1293905182SVP 12 2.927339004 0.1714929166MIMOCC 13 0.6446901580D-01 0.4253584376D-0114FOCC 14 0.3444240583 0.1902614573,DFOCC 15 0.2696233293D-01 0.8365-155980-02DA/PW 16 24.00595430 55.95157687AV'EMPF 17 0.4221124342 0.2803493161FDSL 18 0.3459802204D-01 0.4481312654D-01ATPFAS 19 0.4773171324D-01 0.1667134298D-014flININ 20 0.3475637908 0.1746140554EDS16 21 2.772782533 4.497702891(INXMPF 22 0.2c-'0893884 0.31.81097695XFMOCC 23 0.3, 3803)13 0.1942210881iMINOC 24 0.4358554070 0.1707313442RAXSEX 25 0.3402187120D-01 0.1813956022
0'.1..4C..)0 fr:3
C08REL1TI08 nA78IX
TAVIAdLEitOWIER
1
2
3
4
5
67
8
1
1.00000
2
0.2435561.00000
3
-0.259644D-010.2767801.00000
-0.09110.'3-C2-0.8491710-010.5822340-01.00000
5
-0.140773-0.181035-0.165631-0.8856240-01
1.00000
6
-0.469836-0.1028300.4576770-010.6026190-010.1105350-011.00300
7
-0.855559D-0t-0.130286-0.1510070-0-0.715663D-00.235227
-0.4347300 -C1.000G0
0.7490010-01-0.4215710.1786490.2008000-01
,-0.2279270 -010.1353520-010.614748D-01
1.00000
1.181.18L2 94;mnd
11 12 13 14 15 16
0.211232 -0.1,,S451 0.6573000-01 -0.148,,,67 -0.293755 0.235451 -0.2648160,01 0.3112740-01
0.145712 -.7.4994)10-01 0.4447400-01 -0.4641450 -01 -0.5205443 -01 -0.6307700-01 -0.167245 : 0.230333
3 -0.1090b00-01 0.95/5480-01 -0.4321920-01 0.3933390-01 0.4416040-01 -0.114300 -0.9014110701 0.866676D-01
4 -0.3732890-01 -0.4711540-01 -0.383855D-01 0.1391920-01 0.9422730-02 -0.8384260-04 0.1121810401 -0.277507D-02
5 0.1671550-01 -0.402641D-01 -0.4602820-01 0.3029620-01 0.6717150-01 - 0.4489183-01 0.2067300-01 0.5777530 -01
6 -0.275306 0.7719080-01 0.7472510-01 0.4312443-01 0.173597 -0.121620 -0.4224800-01 -0.2329760-01
7 -0 1178303-01 0.4420493 -01 -0.4666860-01 3,3565680-01 0.136596 -0.8604690-01 0.3941840-01 -0.3647760-01
8 -0.7709033-01 -0.105469 -0.104925 -0.3124570-01 -0.301677 0.435640 0.187836 1/' -0.123942
9 1.00000 -0.5461010-01 -0.1555100-01 -0.9888590-02 -0.114195 0.123937 0.1100840-02 0.107470D-01
10 1.00000 0.2025963 -01 -0.2767130-01 0.446536 -0.395830 -0.232645 -0.9824770-01
11 1.00000 -0.262215 -0.2753410m01 -0:218518 -0.363547 -0.1773050-01
12 1.00000 0.204105 -0.4573500-01 0.448276 0.3730110-01
13 1.00000 -0.641743 0.874430901,02 -0.1145683 -02
141.00300 0.456257 0.1790020-01
15,1.00000 -0.427412D-01
161.00000
a0U
00000'1.
SZ
t0-130POP9f'0
Az
tt-c^t,Cza"0
IT
1C-C199StS*0-
ZZ
LC-COCOPDR.0
IT
tC-C9 0LCcL*0
CZ
1.0-CV'SZtt.'0-
6t
29-c:c.atP*0
St
10-Ct*F6CP7'0-
LI
tc-c(rIt511-0-
91.
IC-CL7t-ft.-0
St
to-co,FrZ9Z*0
hl
1u-c94CW-0
Ct.
70-cttO9tS.0
TI
tf-(707.0..
Lttc-c,,77cf.otw:.
Ot
CO-Cl^ttCQ*C-
6
tf7.Sti'0
O
400(TC0
L
to-ciACt,:*0-
9
t0-Ct9SE9t-0
S
t0-0PtCt,11*0-
t0-n96iZ;f7'0-
1LC.Pk*t*C.,-
1r-cot-s9m.0
C2CUCX
SZ
VISYISTit
006.'0'1
ist
0E0186'0
00000.1
Et
10-0081PW0- 10-0ALS9LE-0-
0000J'1
tz
Z0-Gt6t6CL°0
Zu-Oti.tteL°0
10-0ttZtltS'O-
C0000"1.
tt
t0610F°0
'
PbtZ62°0
616SEE°0- 10-0099SWO
00000.:
OZ
tellS9Cu-
tE9SW0-
969I000
1.0-008EEt8'0-
EtAten*0-
00000't
Ft
SOLOt°0- 10-ObS6L60'0-
99srozo0
t0-0C69tht0-
09Z6SZ*0-
S''S[Pt.0
0000Vt
Pt
EL9901.*0
EPULt°0
hOSCS9°C
t0-Ot?ElSZ7"0
ZOott1-0-
tS96W0
6tILSSI0
00000'1
LI
L0-OPLQSMO
t9-04,169St°0
7s6t[Z°0
10-C2'0901C0
6LE,4Pt*0- 10-0ESZZ0- 10-0tSittS°0
1Zt0Rt"0
Qt
LZ96SS*0
PZ006f) '0
F7tFtt°0- ZO-COZSIS1'0
LOZS,WO
ei1tSP1.0- t0-0007tFL°0-
SSL1191'0-
St
Lt,91.c'0
597666.0
10-0a0tFt°0- To-ocs/66co
0C7t6°C
Lo*Ot°0- t0-0S1Z,PP°0-
t9CC:71'0.
*1
SLSSiabe-
tozervo- To-oarPoo
ZO-CitP66S*0-
Pt5E71°0- CO-OfmStg°0- 10-CL40t:t"0- 10-0,7tS10,00-
Et
10-CLP*,Stt'b
I-001,61)SZ°0- to-chr(rito- t0-019069[°0- LO-OSPOt/t9"0
1.O-ctotc.60- 10-0i4'1ZtS*C- to-cr1 at-90- Zt
C91pc4-0-
7ZL6ZZ°0- 10-0S9St1S'0- t0-041(164-0- 10-491(opnt'0-
tO-Ottr(07"0- 10-Citcp,,,rt.0- to_ccl.rcLie_
It
^Si:int-0-
10/L6E-0- to-OrICOR-S°0 .10-0761t0-
*Thortt C-
7,017W0
SttOtt'D
10-0ScI1WO-
Ot
9t/6010
Ast.'ziol
To-crsziro
to-crnhcco-qo
1.o-c59Pktz.c- 10-naffttl."0- t0-0t,tZLF'0
10-CLTL,,71*C
6
Ot SFI tO
6hAtqh*C1
SeZ091'6- 10-CtPtiPt°0
St^0Stt°0
6r6itt°0- t0-CePtCL.0- 1.0-19Sc4W0-
9
t0-06S766S°0- 10-0016SZ14-0- 10-089.779(0-
ZOZtql*0
t0-CiROI,Pf-O- to-aS459sz-0- t0-eel/v.1C0
tt-c,ctr«c-0-
L
10-00tcEp6°0-
1960710- 10-C6ttiOi"0
t0-01.6bLF6*0- 10-0,16tf,..h*0
tfLttt'0- to-CAS6ttt,'D- 10-tt[t1ft?-0-
9
10-Cf.L9tiC0- 10-0t9L0E9°0- t0-0,1ZSEICO-
16pfr60
t0-05579990
t0-0,4771.6'c- t!)-(1LCZC(t*7,- iC-CtrPt.4tC-
S
Z3-0La03Z*0
rp-ooclohou
10-00PO161°0- 10-07/6tSCO- 10-C9c7nt,C0 10-66"0
10-0719609*9- t0-C,(qi,4*-
t
n6LOZt*0-
C4S,,tt'0-
POIF6SL°0
tfOII°0-
Ctit0./7c-
t6ttici'0
(4CIZI'C
trOf77C
t
10-CSI,S6t6'0- 10-0$,I)6Fp0.0_
OSC9W0
9LttZt*O-
9trtt,1*0-
Sttitt-0
ciGtwt*0
Cf*.t/f*0
t
F6LLPt°0
tItt,6ZZ*0
tO-OSS5fSti0
10-021t4LO°0
10-0t,CSI,t9°0-
toCPW0 tO-C,SCZS.0
tCZsPt°0
'
t
qwnot
Et
ZZ
IZ
OZ
61
at
It
2111118ti
VAR(A8LE
SCHOOLWAGEUNIONMIG<50S. 16hXPTRAINPHV DMNEG WTSVd%BMOCY,WFOCC%eFoccDA/PWAVEmPFFOSLpiPI:ASZPININ[psi('1VvXPFFVOCC
%MPIOC
A
363
Occupation Stratum 6-9 White Males
- __MEAN STANDARD OLVIATION
1 9.590645161 2.6484252712 2.957385305 0.98016378813 0.5591397849 0.49/38233324 0.4597813620 0.49919354275 0.3046594987 0.46109013256 29.75985663 11.936117497 0.1541218638 0.36171434268 3.043010753 0.7621530251
1.161250323 1.20212'0910 4./87313620 1.15031832711 0..3205/347670-01 0.46425196370-0112 0.57648745521) -101 0.124091048913 0.49569892470-02 0.78935',97350-0214 32.3337572 77.18030955
15 0.3/32497013 0.2682779131
16 0.29120546440-01 0.35846870)90-D117 0.44227518570-01 0.16032933450-0118 0.2515325161 0.14083?967719 2.63082430 4.241358082
20 0.2.4547264C4 0.30012138)21 0.62605734710 -01 0.130229358122 0.1446630824 0.1224613621
C1,1AtL1TICN 1ATP1X
vA21A0LEMd,88,1
1 2 3 4 5 6 7 8
1.00000 J.25542.7 0.9945110-0 0.3440240-01 -0.216698 -0.450620 0.214152 -0.7070800-02
2 1.0.Mbo 0.423892 0.363224J-01 -0.1'5)97 "-0.118761 0.100352 0.4519250-01
1 1.00000 0.160062 -0. 06482 -0.1466740-01 0.391280)-01 0.9764540-014 70000 -0.109339 -0.2913570-01 0.2534810-01 0.4249560-01
S 1.00000 0.2837380-01 -0.131571 0.1570616 1.00000 - 0.233012 -0.5105490-017 1.00000 0.1501230 -01
8 1.00000
10.4101A81E IC 11 1. 13 14 15 LaNuw,Eu
1 -t..150i213-C,1 0.616100 -01 0.1491260-01 -0.91457v1 01 -0.1125100-02 0.125968 0.7315370-01 0.9328180-01C.100," 4 3.1,11,,t, -0.1,.549.. -0.%481174-01 -0.111.041 0.107939 0.151174 0.1846850-01
3 C.231,1"2 3.12952) -0.119475 0.1fe7610-01 -0.5707100-01 0.3/10420 -01 0.255555 0.4861800-014 -0.3343e50-01 -0.4154540-01 -0.2123323 -01 0.114684 0.2602210 -01 -0.7025160-01 0.1797820 -01 -0.1274690-01
0.177103 -0.2950730-01 -0.110214 -0.104244 -0.127834 0.5823140-02 -0.187443G-01 -0.1006396 C.8493820-01 0.4424470-01 -0.1171640-01 0.7650240 -01 0.7418590-01 -0.103540 -0.3322360-01 -0.8296800-01
-0.2428390-01 0.4516230-01 -0.4415140-01 -0.2403370-01 0.4649750-02 0.6345200-01 -0.3996491-01 -0.3217020-010.604873 0.113431 -0.0044760-01 -0.364054 -0.542603 0.6301910-01 0.9188930-01 -0.1774880-011.00000 0.449626 - 0.318569 -0.170175 -0.339503 0.9267240-01 0.313317 . 0.7481100-01
1.00000 -0.496038 -0.3136110-01 -0.9794830-01 -0.5289620-01 0.1266411' -01 -0.104050II 1.00000 -0.347068 -0.162800 0.133902 0.8465000 -01 -0.1060870-0112 1.00000 0.765068 -0.8816070-01 -0.125579 0.3073900-0113 1.00000 -0.131278 -0.125043 0.9864220-0114 1.00000 0.455112 0.232623IS 1.00000 0.29Y60816 1.00000
VA4LAULE 17 18 19 20 2161,018E R
c .t '4 %I2 -3.:213411) -C2 -r.Ael14c00-01 0.119554 -0.9341730-01 -0.9368980-012 C.,,r',:r..,-,:i -3.1-)12. -o.,,e141e1) -(11 0.4.-05109 -0.6222030-01 -0.1289663 C.111.-, -1.(5...17 -0.143410 0.13s 16 0.9191110 -02 -0.355131-60=014 C. ,4.)40-)7 -1.4450120-01 -C.9351,;.'10-01 0.118812 0.110856 0.1064065 -0.t204c1 -0.114149 0.014767 -0.107073 -0.107079 -0.1556536 -0.4113"5/-02 9.2422,'70-01 -C.415',r10-01 0.2521660-03 0.00:9880-01 0.5711840-017 /t.')2 3.015x5 -01 -1..511550-01 -0.124562 -0.9711000-02 -0.2260690-01 -0.4261500-018 -C.95/4740-01 -1.14:,c7 ...Loin,: C.763a100-01 -0.3197S3 -0.4381629 C.9 "51 5 70-01 -5.155954 0.110,103 0.359667 - 0.132120 -0.315060
1 0 0.,2471 1.3451/6:)-01 -0.7042470 03 0.6614380-01 -0.3583690-01 -0.226161II -0.2Sr5C4J-01 -0.495)15-51 -0.14250 0.5139010-01 -0.340577 0.16920700112 -0.162:773-01 0.458/79 -0.9449570-01 -0.117316 0.909238 0.931048I3 C.110122 0.453632 -0.114581 -0.124959 0.759419 0.77798914 -C.171129 -0.1911E4 0.1701550-01 0.267411 -0.9196240-01 -0.4703350-0115 0.348564 -0.2421/5 -0.1173160-U1 13.720508 -0.127239 -b.10322016 C.172754 -0.153579 -0.9342200-01 0.213956 0.3526910-01 0.3346450-0117 1.00000 0.6649150-11 -C.7942370-01 0.244231 -0.3109240-02 -0.1462600 -0113 1.00000 -0.8234050-01 -0.288295 0.464650 0.47532419 1.00000 -0.7742470-01 -0:9698670 -01 -0.14341820 1.00000 - 0.119360 -0.10744921 1.00000 0.9343192? 1.00000
I
365
Occupation Stratum 6-9 Black Males
yARiAOLE
SCHOOL 1
WAGE 2
UNION 3
M IG <50 4
S. 16 5
6---EXP- T
MEAN
8.8494623662..359516129
0:43548387100.47311827960.698024731228.263440E6--
STA41DAkD,DEVWION
2.9406:)08430.94179029180.49715844690.50062442850.459963482411.86092640
TRAIN 7 0.1075268817 0..3106180619PHY ON 8 2.903225806 0.6666119574NEG hT 9 0.7096/74194 1.153694886SVP 10 4.330645161 0.92723163261:8MCCC 11 0.1076666667 0.3744345942D-01
--%WECC0 12 0.38602150540-01 0.5343015441D-01-.%FFOCC 13 0-51720430110-02 0.662856GU63D-02DA/PW 14 21.80380573 143.42801756AVEMPF 15 0.3162598566 0.2689391596FCSL 16 0.472460(: 94F0 -01 0.1094640047ATPEAS 17 0.39948924730-01 0.16611o8886D-01gv1141N 18 0.2E20625306 0.14;3201:',648
ECS16 19 5.7419354E4 4.470693098UNXMPE 20 0.1923949821 0.289167-'161
21 0.41774193550-01 0.5933',1'6930-01tVINO0 2 0.15144C8602 0.55205351D-01
:1,83
VAKIA6LE
ALItitA
MATRIX
23
45
67
-
I.ccac.3
0.19
4143
0.1.30544
0.3028190-01
-0.329417
'- 0.4596..L
0.207185
0.4214990-01
1.o0oc0
0.4/7591
-0.25''19
-J.274858
-0.2340930-01
0.3639519-01
0.0000450-01
31.00000
-0.22
7-0.274505
0.124357
0.1016221-01
-0.51:4590-01
41.1
JO
-0.8228030-01
-0.9119910-01
-0.508330)-01
0.8931970-01
51.00000
0.100817
-0.112087
0.1023323-01
61.00000
-0.216070
-0.130032
71.00000
0.180998
a1.00000
v6k1
.1.8
1(61
.' "E
A10
1112
13
14
15
16
1-0.4743'01-v1
-^?.2(.,G,.154.6-01
-C.1),a8e1
6.81v:16J-01
0.6501180-01
0.8606180-01
0.1443120-01
-0.3259110-01
2.
C.112:,..1
3..11412
-6.159510
6.874404J-01
0.4109k.50-01
0.5102720-C1
0.346695
0.112910
3C.1111S9
0.2757e6
-C./21436
0.4105280-01
0.5195410-01
-0.3734770-02
0.411086.
0.9221843-01
4-C.1011S6
-0.193264
0.111464
-0.53454,0 -01
-0.11[022
-0.1849120-01
-0.8012690 -01
-0.189384
S-C.5355060-01
3.5274C67 -02
0.5317450-01
-0.5148713-01
- 0.5915410-01
0.2914030-01
-0.125154
0.2159280-01
6C.3721460-01
-0.1428650-01
0.9441140-01
-0.2210420-01
-0.1130500-01
- 0.7816880 -02
0.030004)-01
0.47/9000-01
7C.102649
0.104858
-0.120591
-0.7362183-01
-0.119247
-0.4276000-01
0.6133331-01
0.8070370 -01
8C.412900
0.205000
-0.4115610-02
-0.303238
-0.594231
0.326716U-01
0.8408910-02
0.6297580-01
91.00000
0.634824
-0.404.942
-0.9876440-01
- 0.250618
-0.2942640-01
0.339992
0.254176
10
1.00000
-0.503993
0.3542360-01
-0.7245130 -01
-0.3011910-01
0.239844
0.349893
11
1.00000
-0.426921
-0.334059
0.139733
-0.209031
-0.225793
12
1.00000
0.818713
-0.162028
- 0.217329) -C1
-0.8707820-01
13
1.001100
-0.166115'
-0.425247)-01
-0.151410
14
1.00000
0.333302
-0.3723550-01
15
1.00000
0.138210
16
1.00000
N/A.1Antt
C19
20
21
'22
M..$06ER
1c.se45490 -01
6r.J-01
0.1,5738
0.5710400-01
0.0408160-01
-0.1906690-01
2c .2
7(44
3-).131105
-0.15..509
0.456465
0.8333030-31
-0.1856530-01
38.5,)81.4
8.5,)81.4
-3.1234E3
0.159569
0.4405891)-01
-0.9940170-01
4-C.2346660-01
0.60ril2p-q1:=1
-0.107488
-0.8118110-01
0.9819400-02
5-C.737)66
-1.9 34710-01
0.845231
-0.176426
-0.5035290-01
-0.2667570-01
6C.47c31.6,)-31
-0.61.321.3-01
-C.94w227o-1
0.121921
-0.2116590 -01
0.4430080-01
70.1919/1)-01
0.1.7451
-0.4218950-71
0.2134100-01
-0.7961670-01
-0.167327
>-0.1946700-01
-3.2c7111
0.4)8),,,0 -01
0.3282950-02
-0.393440
-0.425093
9C.110611
-J.16a139
-0.1451450-31
0.253422
-0.110932
-0.400196
10
C.1154.15
-3.110455
0.157'.01D-J1
0.230517
0.23a020.0 -61
-0.316069
11
-C.17e213
-1.6145C50-1
0.1311733 -01 -0.137822
-0.421948
0.224351
12
C.2C641.
9.52tAts
-0.1701170-01 - 0.1129593 -01
0.9,4577
0.783943
13
0.195624
0.422345
- C.'75(.040 -01 -0.4230870-01
0.932917
0.742115
14
-C.73(f1130-01
-311169759
0.82060-01
Q.189379
-0.1646/3
-0.8227900-01
15
C.52t126
- C.2430'6
-0.8864140-01
0.76/123
-0.2431900-01
-0.168304
to
-0.246925
-3.216967
-0.8600133-02
.0.096,040-01
-0.9532040-01
-0.255522
17
1.00000
-0.3989690-01
-0.145557
0.506886
0.237758
0.102505
18
1.00000
-0.7546433 -01'-0.259967
0.338791
0.322788
19
1.00000
-0.116704
-0.3917910-01
-0.3322090-01
23
1.03000
-0.2390140-01
-0.153015
21
1.00000
0.768844
22
1.00000
a rY O
367
Occupation Stratum 6-9 White Females
vA/TAltF
crqint7
E AN
n.16949n566.1.°3f,'037740.4'330672642
STA(,!0A20 DEVIATION
7.66757547B0.4R072541771.49717413.33
"fle90 (1.'49°679745 0.c079174931
C. t4 r. '1.749?n3n189 0.4122913551ryr). .31.90041396 10.74941753TDAlm 7 0.7win1rFA790-91 0.16/4,217370
WiN0 0" 7.1017,1595 0.A111901756'on wr 0.'(.41g00414 1.777156886SVP In "3.19(10413/6 0.7814915335conAncr,
m!W="Cr.
11
1
n.11r:6(-n1774D-11n. A$111(1/0157
1.7910161777D-010.317316x414
°"Icr1CC 13 0.3780672642D-01 1.19753503470-01nApw 14 9.61)276917 16.9391:i156
0.1617443196 0.1907(130512
1, 14 r).?,132qconsin-01 0.1070 -160047D-01
ATPFAs n.,:i44r8A7070-01 0.1?7)9-190130-91'%iTmIN1 0.61173320?7 0.1809724475
cncl4 In 9.41,'264151 4.41114202Sowc4or /n 0.5(177'10037D-01 1.12'7317796m,rk4ncr 0.77612.11321 0.1925912933Tmtmnr 27 0.7461471609 0.12/8800414
(191-t11109 !Tot(
vall3niePi.grO
1
1
S7
1
1.1W,10
7
-3.1461421.0npoo
1
-9.1483040.1169271.00900
4
1.1)1143-3.200216-0.2997401.00000
5
0.146181-0.119773- 0.3664450.2235610-011.00000
6
-0.3502100.4679540-010.299936
-0.1067340-01-0.244412
1.00000
7
0.1057160.117945- 0.1494310.5046390-0t0.349250-01
-0.9853040-011.00000
8
0.2298130-010.6192593-020.6858230-02
-0.111399-0.9414340-010.8475760-01
-0.116711
Vic!IVALF4,1",r2
A
5
7a9
111!
121114IS
".117,7 10_11
-11
-''.1 )97,°7,_11.0.c179410-11n.o117^94-11
0.7441541.9'1119
In
1.161701-1.1641010-1?-1.1:5442-1.4174771-011.°14a100-01'.4Q1,..673-011.1474191.4716561.9560,4wyleno
11
0.,714551.4774700-01
-9.211,260.4481420-01-0.4641950-01-0.741491n.1192730.3118660.1891780."91576I. 00900
I?
-0.?77174-6.41n837)-01
=1481-010.6510620-010.706962
-0.784550-0.445292-0.640928-3.577357-0.8913651.00000
13
-0.1'1619-0.7166150-010.171127-0.4642950-010.5485850-010.151553-0.148197-0.345162-0.629266-0.388618-C.6700280.8051371.00000
14
00:11=-0.144582-0.7225150-010.141742-0.2242550.103910
-0.1260000.6932340-010.1688690.5801890-01
-0.9492740-01-0.244428
1.00000
15
0.2102430.150111
-0.2485100.3179720-010.1655210-01
-U.1631840.2378190.3519580.6355460.5217200.587862
- 0.748560-0.6504990.4969001.00000
1.0000016
-0.1911920.1129780.152245
-0.114568-0.1256270.101601
-0.2040710.2042970.171829-0.8121693-01-0.2779340.2168960.3128910 -01-0.6392960 -01-0.267609
/ADItAIFNormc*
17 18 20 21 22
1n.,C1.^7 -1.1491,4 9.7671.41 -0.6111570-01 -0.223997 -0.216615
9.19&7n19-11 -1.Inn'014 -0-1(4797 0.110159 -0.4440240-01 -0.349:020-01
1 -'."1151 1.271"1 -0.111SiQ 0.4608)8 0.230136 0.233483
e. n.114,11 , -1.651A/41-11 0.715944n-91 -n.182168 -0.4581960-01 -0.4545430-01
S -n.,90,419-01 -1.119°4" 0.967116 0.6107361-01 0.6845250-01
_90 .-4,,4 1.117091 -0.116105 3.189973 0.236057 0.197461
7 0194975 -9.1c,160% n.6411165,7-01 -').7920763-01 -0.740115 -0.272149
4 0. -./. /-,/ 2 9 -1.27c/21 -0.9117940-n1 0.360256 -0.446530 -0.451740
n 0.'.44447 -1.470744 ..0.414,.1,0-,1 0.507513 -0.646153 -0.661478
11 6.1418,8 -1.471,80 0.1 n}17 0.316471 -0.575185 -0.539777
11 9c...9949 -1.371050 -n.1347390-11 0.201350 -0.848388 -0.869636,
1) _n.474116 9.4970/.9 0.4273850-01 -/.101171 0.1)9)12 0.998435
11 -n.c184,17 1.5517c,3 0.cS51543-01 - 0.341184 0.823445 0. 28079
14 r.110941 -1.410471 0.21149n 7.1537081-01 -0.111816 -0.106667n.4,415q -1.597P54 n.4117811-01 0.1634750 -0.751711 -0.756927
t. -n0990114 1.12134? -0.117069 0.2091873-01 0.201164 0.201107
1.-11119 -1.643499 -0.1865270-01 1.111858 -0.434170 -0.642637
18 1.00o00 -0.206564 -0.274215 0.514303 0.521274
1.01000 -1.174954 0.4341570-06 0.4553290-011.30000 -0.306464 -0.312347
71 1.00000 0.9992261.00000
1.60000. _
VARIABLE
369
Occupation Stratum 6-9 Black Females
MEAN STANDARD DEVIATION
SCHOOLWAGEUNIONMIG<50
1
234
10.116279071.721860465
0.39534683720.4418604651
2.3725544030.47812259740.49471179120.5024855166
16 5 0.6279069767 0.4890834876.S.EXP 6 28.37209302 12.09687731TRAIN 7 0.2325581395 0.4274625744PHY DM 8 2.162790698 0.6521113168NEG WT 9 0.2325581395 0.6109013365SVP 10 4.372093023 0.8247151258UMOCC 11 0.2511627907D-01 0.3197857118D-01'AWFOCC 12 0.6157441860 0.3513592039%8FOCC 13 0.4020930233D-01 0.3433487616D-01DA/PW 14 4.801214906 5.686813679AVE3PF 15 0.1333108527 0.1353513931FDSL 16. 0.2840904454D-01 0.3100309167D-01ATPFAS 17 0.3459767442D-01. 0.1141368202D-01UlININ 18 0.5865936475 0.2011410698EDS16 19 6.232558140 5.277252213UNXMPF 20 0.6325813953D-01 0.1258360268ItFOOCC 21 0.6559534884 0.3673662992%MINOC 22 0.6810697674 0.3403461400
S7
.'CORRELATION WHIZ
VARIABLE 1
sussza
1
2
345
78
VARIABLEsus888
1
23
456
89
10111213
1516
VARIABLEsumm
1
23
4
5
6
7
8910111213141516171s19202122
1.00000
9
2 3
0.2462181.00000
4 5
-0.4009690-01 -0.104038 -0.1054570.503139 -0.106571 -0.3635171.00300 -0.4900370-01 - 0.263175
1.00000 -0.2838871.00000
10 11 12 1)
-0.150519 0.112430 0.221056 -0.295434 -0.356303
0.335958 0.442916 0.202690 -3.341877 -0.477512
0.161228 0.384500 -0.2103510-01 -0.7145410-01 -0.167589
-0.187595 -0.6146283-01 -0.401858 0.351555 0.3067320-01
-0.101927 -C.250666 0.9417200-01 -0.4324150-01 0.6855170-01
0.6655813-01 -0.2399330-01 -0.6246330-01 0.7589340-01 0.278693
0.614939D-01 - 0.129736 0.344589 -3.281454 -0.199607
0.7394371.00000
0.424801n,5697361.00000
17 18
0.3336030.2806490.1388221.00000
-0.375415 -0.4834730-01-0.525943 -0.135186-0.373842 -0.20851.18
-0.866494 -0.3046251.00000 0.427965
1.00000
19 20
6
-0.576446-0.1565700.1331750.7560730-020.1487131.00000
14
0.2)12020.4631850.191797(1.1090663 -01
0.1045950-02-0.2092870.2188120.2366850.512)020.4645730.167727- 0.416108- 0.313316
1.00000
21 22
0.357159 -0.251765 0.245001 0.149282 - 0.313862 -0.320168
0.595577 -0.371767 -0.284964 0.471517 -0.371610 -0.382067
7 8
0.183992 -0.1048610.115494 0.257880
-0.107353 0.9096710-01-0.157250 -0.2247440.8210370 )1 -0.106167
-0.307214 0.1628450-011.00000 0.3178210 -01
1.00000
15 16
0.1556990.4131000.1613670.265220D-0
-0.171836-0.1115610.2629430.1507160.3015120.3544460.478609
-0.634201-0.3860780.5605701.00000
0.292465D-01-0.1237380.3486870.202748-0.291625-0.152559- 0.137345- 0.185211- 0.261505- 0.217022- 0.2809130.3234730.2639730-01- 0.141276- 0.138079D-01
1.00000
0.532314 -0.271119 -0.236694 0.624037 -0.8400330-01 -0.9264930-01
-0.130173 0.219195 -0.326996 -0.5806480-01 0.339852 0.329074
-0.204036 0.7426340-01 0.919912 -0.313065 -0.3495030 -C1 -0.2887673 -01
-0.194613 0.1163460 -01 -0.4d7565D-01 -0.6214546-01 0.9663380-01 0.100595
-0.5893550-01 -0.9183020-01 0.133774 -0.9891920-02 -0.287854 -0.273329
0.247,68 -0.321029 -0.170392 0.154607 -0.367601 - 0.365225
0.409845 -0.562535 -0.201809 0.340913 -0.515661 -0.521176
0.548194 -0.55:,578 -0.256141 0.537394 -0.377'046 -0.396755
0.349061 -0.2h5848 0.143038 0.233327 -0.057397 -0.831507
-0.544160 0.553492 -0.117422 -3.422437 0.996426 0.9)4118
-0.559519 0.6729490-01 -0.7530650-01 -0.260153 0.502779 0.513385
0.533105 -0.517700 0.3508590-01 0.316964 -0.427232 -0.445391
0.626880 -0.434050 -0.171909 3.663343 - 0.642838 -0.646403
0.14o039 0.136249 -0.259152 0.295505 0.312032' 0.310410
1.00000 -0.599430 -0.115060 0.719715 -0.572763 - 0.585437
1.00000 0.349300-01 -0.494802 0.540446 0.556494
1.00000 -0.358457 -0.119344 - 0.115379
1.00000 -0.428345 -0.4434281.00000 0.998830
1.00300
371
Occupation Stratum 6-9 Cross Race-Sex
v/01ARLF
Sr1-11-.11 1
w1 ,r3
m1Ge51 4.
S 16 6
6
MEAN
9.5130023642.6('07801420.52nr9456260.4860976359n.311146916229.96690307-
STANDAPD DEVIAT104.
2.6357362200.99920376190.50018763360.50042277620.466698837311.51194894--
p AGE 7 0.89P14515370-01 0.2862831.281
SFX 8 0.26713947q9 0.44298973160.1276595745 0.3341055222
PAY 04 11 7.787734043 0.8215853704MFG, vT 11 0.9010732861 1.191438200SVP 1? 4.A4i626478 1.071595136Y.rl'Ancr, 13 0.6733P96927D-01 0.51158150430-01TFlcC 14 0.2216312057 0.3426666837ev.111CC 16 0.1?510638300-01 0.11524296150-010A/Pw 15 21.1f-4)5290R 64.49889744Al/PmDF 17 0. 31812 ?2 ?22 0.2657181447FOSS 1q- 0,2043917050D-01-- 0.3955346660-01--ATPFAS 19 0.4162624113D-01 0.15522860280-01.wmiN1m 21 0.3568197721 0.224 827782Fns16 21 2.891257955 4.479232916liNvinp 2? 0.100(1367218 0.2715906152`1'F '1C(: 23 0.2351418440 0,359822084414r/V1r. 24 0.3024728132 0.3281326003RAXSEX 26 0.16548463360-01 0.1277230486
X89
C",
99cl
&T
IOn
mAT91x
VaqIARIF
I
por.qc*
7A
45
67
8
11.109)1
1.156100
-0.2311590-01
0.5447530-01 -0.110003
- 0.424602
-0.7032530-02 - 0.1820320 -0!
7-0.4900571-01 -0.157169
0.411313
-0.0722780-01 -0.154321
1.00n0n
..
11.00000 --
'
0.2708090-01 -0.255140
-''
0.5608440-01 -0.2918480-01
- 0.463264
-
1.00000
4-0
.481
4160
-01
-0.2969140-01
0.7433550-01
0.3113990-01
61.00000
-0.6330680-01
21)0:11626h
-0.1,804120-01
61.00000
120-01
0.6028630 -01
1.00000
-0.5888270-01
41.00000
16
vaotaAir
010
11
12
IA
1415
vp...,1-0
11.Ingli9
20.1,0'244
11.7716/90-01
40.741745n-01
S.41.4434510...11
4-0.716716
70.7401391-11
6-0
.134
699
9I . A no1 n
10
11
-......
1713
14ISI&
10117170-91
1.7619?7
0.1.70,01
-0.6744700-02
1.101153
-1.4953300-01
1.109)230 -0l
-3.007c73
0.5602110-01
1.00000
--.
""
-1.11458.1U-11
(rA,T41v
.'14 4
-a.3520560-01
0.116405
0.7381210 -31
-0.1024410-01
-0.330740
0.1941110-01
0.652506
1.00000 "---
00-,04790-01
0.5)11760-01
3.227729
0.104611
=0-01
-3.55,1789-01 -0.17,41110-01
-0.12601410-01 :0.1140150-01
0.2075333 -01 -0.110649
-3.6514550-01
0.163001
-0.171919
-0.460188
0.7966390-01
0.5561250-01
0.217413
0.265253
0.510369 "---=0.5623460-02
1.00000
-0.242591
1.00000
-0.104683
-0.1496.)1
-0.213959D-01
0.4594180-01
-0.6502310-01
0.118268
-0.5176720-01
0.806223
-0.161241
-0.575252
-0.410970 '-----0.462026-----
-0.238545
-0.664956
1.00000
-0.5706670-01
-0.1855S2
-U.5d19tIOD-01
0.1401010-01
-0.8064240-01-
0.108879
-0.8065500-01
0.776591
-0.135377
- 0.611055
-0.236240
-0.582745
0.922260
1.00000
0.115070
0.171715
0.3612310-01
-0.6966020-01
0.1728610 -0/
-0.100289
-0.5056270-01
- 0.170181
.0-01
0.140835
0.140630
"'-
-0.7990010-03
0.186273
-0.168414
-0.193632
1.00000
yA41A9LF 17NUmArq
In 14 20 21 22 23 24
%)...111c0-11 1.17'15"-11 1.1."613 -o.4 )1/510-11 0./3)4170-01 0.87002)10-01 -0.101630I, nVW.01 1.11"r0-01 0.)111,1l
n.t11179-1.41t94; 0.443244 -0.3918/6
-0.143,696-0.4u/060
1 ^.19s419 0. -0.110'63 :1°3.2= 0.658390 -0.2399660-01 -0.3101690-014 -0.15S1170-)7 -.1.6468150-0.) 0.2499090-01 0.2091143-02 -0.1214120-01 0.4229460-01 0.4449160-01 0.4600060-01S -0.1/21s90-11 -1.6174400-01 -0.1175/1 0.941669 -0.66846/0-01 -0.1509250-01-0.11s738s ..0.ss06.110-01 -0.6206sen-01 -0.16'16010-01 0.7660060-01 -0.137137 1.11=001 0.118235 0.1124027 -m.111,1100-'1 1.711411D-11 -r.144710-01 1:1=9D-01 9.216219 -0.1943460-01 -0.9195610-01 -0.7512390-01R -1.114114 -1.1979,40-91 -0.71704? -0.1198,?0-01 -0.2164179 9.4s4le41-01 9.,e11910 -01 0.140121
0.819669-3.5463630-0i 0.3927920-01 -0.160522
0.600526-0.146773 -0.16/355
11 0.761511 1.501'914-J? 0.0901491-01 -0.550601 0.0446940-01 0.213104 -0.571593 -0.59421211 1.,.1.1044 1.972194o-0i 0.'13327 -0.460559 0.105414 0.410192 -0.415162 -0.45613317 0.1544,1 -1.s4it060-02 0.125115 -0.190,143 0.1841360-01 0.142576 -0.239314 -J.30026911 1.271244
- -0.N151670 -31 0.141617 -0.468112 -0.3310/10-01 0.192044 -0.663254 -0.57140114 -0.414850 0.2151520-91 -0.117699 0.770420 -0.3543760-01 -0.30/453 0.999802 0.9926d715 -0.411155 9.2165710-01 -0.746577 0.771067 -0.4433700-01 -0.104053 0.929711 0.92d/10Is 0.456113 1.1c0458 -0.9430570-01 -0.247921 0.2242280-01 0.206328 -0.170353 -0.157764
. _ ...---.17 1.00000 0.180105 0.466049 -0.474520 -----0.3128840-01 0.713001' --0.435388 -0.43514610 1.00100 0.5074310-01 -0.5107653-01 -0.99do770.-01 0.100752 0.2152d10-01 0.174567D-0119 1.00000 -0.266564 -0.9526231 -0.315350-01 0.303897 -0.323362
CO 29 1.00000 -0.751917-01 -0.395977 0.773384 0.774966
CIO
1P
,I 1.00000 -0.1159031.00000
-0.360306D-01 -0.4423540-012? ..0.308443 -.0.30d29571. 1e00000- 0.9931.7074
9341401c 24
1.00000
'M.-1;A
0q7:5,y)-117 -e.:^7,q,1 -0.,170170 -nt4S 0.1190486 -0.54422S0-917 0.412494
r.?14044"%116169
10 -0.1,444,..0.01,a051-11
12 _1.45711,0 -11)3 -4-.975)&99-11
14 4-.:7019is 0.1N44i9Is -0.1.17e710-0117 -1-.41119s0-01le -0019494,0-9119 .4.4795210-917A 0.:7111421 0.11,14:412 -,4491159-0171 n.121113724 0,117qR5
I.A0ft10
n0ffr
374
Occupation Stratum 12-14 White Males
VA: IALILE
SCift-;01 1
WAGE 2
UNION 3
MIG<50 4
1:).i302670623.4974626C8
0.44J652819J0.41246290E0
STAIWARO OEVIATION
2.3494032601.123903964
0.4901)3412700.492412204
S. 16. 5 0.247/7448(7 0.432040/734
-EXP------ '23.40534451-- 11.60850533--
TRAIN 7 0.3234421365 0.46d1372312
PhY 014 8 2.66468427 0.1-J1663316
NH; WT 9 0.92q7833828 0.914(134749
SVP 10 /.011369436 0.1419310./00
F4IIICC 11 0.667655/1i60-01 0.,!02129t.3110-01-
";',..P"CC- 12 0.4/G49E516:10-01 0.7')93s0(=.4480-01--
Y.:I:ACC 13 0.35J14:66:300-02 0.1:)032215830-01
DA/P'e, 14 SE.44435514 7V.62230237
AW:HPF 15 0.4556345657 0.315816796FrSL 16 0. '[:4337561:771) -01 0.30.;56;04710-01
AlPFAS 11 0.44460320470-01 0.17 891916111) -01
18- 0.'2812893636 0.15/531i:641;-- - ----
rosi6 19 2.53/0';1980 4.50)534737
0NY.:TF 20 0.2430857072 0.3450()33416
".1F-11:CC 21 0.11151335310-01 0.12249)55121) -01
V.ANCC 22 0.771327693130-01 0.93650007480-01
392
00000'T P9..0916'0
4SSZ22'0- 70-01 v4197'0-
00000'1 ISI4C2*0- 70-0.,7,9.27f'0-
00000'7 IO-C.,ZCE09'0- 00000'1
22 IZ OZ 61
'7126.7c*0 L.,Q.,1£0 10-04601S4'0- 20-0002.701'0- 011^0'1 PI
9c7.9170- $70917'0- 99751'3 . 13Z4PT12- 10-0E,OSOZ'C .
0(:'0)0.1 LT
ciA7710- 946771'e- 70-046.A911 Z0-2/01(s0 Iti.pc....v- 1271'0- ..1
f5h4.77'0- IISZTO- 67£119'0 C0-0cc...40- 10-nat544C- 151712D 41
Icqc01'0- Wn9'0'0
70-00./4116'0- 964f09'0
7910I'0 sw.0)16-
10-c4,41c4'0 10-0M4,41'0-
r:f.iLl3- 21:11.t'0
c.ft:;;:: r1
710946'0 040766'0 691.6.74'0- 10-('4144.C'0- 1.49 .,tt'0 1:etql*3- 27
599911'0 t0 -ti1 .7916c0- 10-Off15i£'0 10-01:4501'3 TO-00SsuT62- ZLP,ST.:- TI
SP/112'0- .76c,1'0- 994SE2'0- f0-:11'1',6t,,'0- li91Z1,- 9E1..102'3 01
20-01694,fl0 70-06116,1'0 Z.d...e1'0- Io-t:,,L.711')': 70-0!,1:141.1.- .p411.1e.)- 6 c41c710- .7v/10.70- p4qq.2.7',$ 70-0v1.7N.,:"- 19-C1.el,,,'1%- 10-00.147"1,- P
To-oE5,5pi*0- TO-0eZt.:69'0- lo-egeotst.r., 76-0.5..-1(9'L TC-0/IC(.15t. 1C-Or.i,-aT'2- 1
10-016410.,O 70-r4z047ro in-(f pof.,-n- in-ol.10.t...o- 1J-C444.:1C7'0- 10-0ross.,, 9
2o-asr5e91.0- lo-o....,4.4610- 10-016.7119'0- tlt.6560 ZO-09.21::29'e 119/0.:1'0- s
zo-ocsts.,r'n ozse6r.o- .291671'0-
70-06.7:'0,1'0- 6,7s.,er.c- 10-010.7911*0-
1n- (170.075 0- lo-nssts.to- en-vr:..tvis.c .,te..6/13 12/7C1'0- 10-!:9.14.0- 10-()407,,5"0 1.10,1'0- 11'0::1'r-
10-n=:3- 72.611')
;,,
2
100071'0- 10-0&r.9040- 10-(Ph. /1,..u- 1C-Ci.z.0.-1, 10-11.,7Pt2'1- 10-1!!$:..../.3 1
t13t11:11P1
22 '12 Ve 61 WI AI 4101,1bVA
00000'T ! 91
TO-ORM,O 1.'0000'7 St
20-0569691*0- 196.76.20 00000'1 .,I
70-021:9714'0- I7.2141'0- 70-0100/75*0- 00000'7 C1
9796210- .P74r17.'0- 6f2201'0- /0997L'0 00003'1 Z1
10-02£09/10- 107001'3- 10-0.7f,,92E*0- I.:6,492'0 10-0.62166'0- 00000'1 11
20-nolts.,c0 1','c14213- rotwo- 67c117t0- P5.)671.0- 9.,94c7'0- 01700'7 07
70-n700re1.0- 70- IOtiO7'0- 70-0455/CP0 10-0561016'0 20-0ffl.'21'0- 10-01206/L0- 70k179501'0 cGenI 6
70-0050211'0 tt14c.7'0 4f061'.1'0 FO-CPSsOg9'0 PCW0- 16.7q s, 7701cC- 10-0qE..23 9
10-05.29,44*0- totot)vo lo-ose.,slro 70-019516.70- 70-Ca1I:10V°- 70-0Pc120i., u- 10-0.71111.1'0 lo-rti.vto t
70-07990f7'0 70- ;4+4912'0- 70-n101vsue- 79-(1.7(..,20(3.0 10-01/?.,.,z.0 10-01wtzfEeo 7p-or,114/0.0 z2_nctegzso, o
' 20 -W/45/6.20 /0- 41712"0- 70-0609.'15'0 70-0117112-e 70-M693497'0- 10-066r1o1.0 rc-citt,:(Ae.c- 10-c9/p2c6-3 s _ 10-01:9410I0- TO- !10071'0- Z0-0106/11'0- 10-alst.r610 7o-vver.,:,4'0- 10-04wAt.O lo-v4sI71 .,0- l'-vc6+0,ts.:1- ...
10-o2I1261.0 k7.,f,n0 1/.21o1.0 10-0b29f.6l0- 0711.'2'0- 6,-1,1.4 10-0ov"f_ 14e1,1.7,_ c
10-011O6000 i(-00re, 10-n6976,6.n .117.c1.0- 79-rcq4/sso- :ti/?.0- -..1:,.,I.v Itrf./.;', :1 2
10-0C14?46,'0- 10- 0/1,160 10-nanC66'70 5.6017'0- 10-mTOtl0- ".,0,-..t :)... 47'. r.I 1 1 10-f !.il. i5. *,- I
91
00000'7 10-00S6SWO
10-024 7762'0 10-06cc,0E0 70 20-41700ov1.0 sm.pmro TO
511001'0- 10 691'1'7'0-
P
41
00000.1 9F.701C0- ;7201141'0- I59t.,10- 1/0./420- .c..7010 611541'0
41
00000'1 2o-99420E11..0 10-106F.70/0 Z0-07091P.,C 10-0000604*0- ES.,r1.20-
(I
onrowl 10-01070540- £252.;1'0- 2.274:W0- I/9/11'0-
71
oocovi 10-000.9EtE0
6P9 ri1 '.:,-
70-Vc/940-
11
00000'1 .....
10 -016 / 675'0 10-0LC7W0-
_ . ... 0nn00'1
61C,711(
6
0000..01
v 1
9 s
6 t 7
1
P2m10%
I 9 S r 7 1 41..y 14:6
x1 Alva Ne.111:71i,.:.;:3
376
Occupation Stratum 12-14 Black Males
VARIABLE MEAN STANDARD DEVIATION
SCHOOL 1 9.187500000 3.262166808WAGE 2 2.588046675 0.9999307524UNION 3 0.3671875000 0.4839322812MIG<50 4 0.3593750000 0.4817026070S. 16 5 0.6406250000 .0.4817026070
EXP 6 28.01562500 12.57480569
TRAIN 7 0.1796875000 0.3854355089PHY DA 8 2.6)0625030 0.6668101880
NEG WT 9 1.132812500 0.9338017607SYP 10 6.768281250 0.5215997642tBMOCC 11 0.4792187500D-01 0.432222,i637D-01
YIWFOCC 12 0.9700593750D-01 0.1729455801FOCC 13 0.2215625000D-01 0.5031623421D-01
DA/P4 14 17.25026181 32.05703116
AVEMPF 15 0.3624205729 0.3017795463FDSL 16 0.6339899184D -01 0.1348131672ATPFAS 17 0.361226562-,D-01 0.2092413072D-01
XiININ 18 0.3455067623 0.1842227408EDS16 19 5.492187500 4.962847947
UN an 20 0.1947591146 0.3168350711
%FMOCC 21 0.1192421875 0.2215)18084
%1INOC 22 0.1671640625 0.2345855376
394
CORRELATION RAM/
VARIABLE 1
MUSSER2 3 3 5 6 7 8
1 1.00000 0.330818 0.6577610-01 -0.188533 -0.252421 -0.621415 0.,86112 -0.9547300-01
2 1.3000D 0.395277 -0.253223 -0.200089 -0.219274 0.322695 -0.4756010-01
3 1.00000 -3.6386130-01 -0.7125030-01 -3.7341020-01 0.2341580-01 0.3431410-02
4 1.00000 -0.8377520 -01 0.1466470-01 -0.180904 -0.7277610-01
5 1.00000 0.225820 -0.158374 0.146310
61.00303 -0.265392 0.104441
71.00000 0.4643380-01
81.00000
VARIABLEMUSBER
9 10 11 12 13 14 15 16
1 -0.189179 0.7210330-01 -0.126942 3.7493840-02 -0.5966440-01 0.7008760-01 0.264167 0.5124510 -01
2 -0.105299 0.108140 -0.130760 -0.203510 -0.255547 0.141409 0.500351 0.159135
3 -0.126141 -0.7640150-01 0.9172450-01 -0.186566 ...0.176996 0.3314050-01 0.425693 -0.5223140-01
4 -0.3692470 -01 -0.4391760-01 0.5/41010-02 0.5992790 -01 0.6361350-01 0.7135200-01 -0.2998710-01 0.1190790-01
5 -0.3309540-01 - 0.4823210 -01 0.31E11250-01 -0.7552310-01 -0.6426110-01 0.4495540-01 -0.2418950-01 0.2735390-01
6 0.103089 -0.3453280-02 0.2271830-01 -0.5029130-01 - 0.6064490 -02 0.6243280 -01 -0.137441 -0.3725260-02
7 -0.198090 0.131970 . . 9 8 -0.195126 -0.5452820-01 0.199963 0.132066
8 0.2351290-01 -0.325188 0.137942 -0.8035000-01 0.5895000-01 0.173120 0.7964500-01 0.291808
9 1.00000 -0.169756 0.7536850-01 0.339030 0.369414 -0.4231350-01 -0.254942 0.3759650-01
1011
1.00000 -0.681625 -0.520219 -0.605071 -0.4445450-02 0.1489790-01 0.1019410-01
1.00000 0.195638 0.269986 -0.128173 -0.135897 -0.107903...
12 1.00300 0.955022 -0.178870 -0.345307 ,-0.196933
13 1.00000 0.181918 -0.349505 -0.167377
1415
1.00300 0.545967 -0.9085410-01
16
1.00000 0.120956
VAI'IlBLI 17RINSER
18 19 20 21_,
22__ _1.00000___
1 0.216900 0.4552190-01 0.257853 0.137372 -0.7702240-02 -0.1065170-01
2 0.378133 -0.319145 -0.2463660-01 0.432375 -0.216943 -0.229)40
3 0.4230S0 -0.172243 -0.1682810-01 0.810142 -0.185675 -0.156607
4 -0.5549913 -01 0.110651 -0.161502 -0.3408970-01 0.6124130-01 0.5077290-01
5 -0.126289 -0.132899 . 0.832127 -0.6956590-01 - 0.7356650 -01 - 0.6323250 -01
6 - 0.231511 -0.3937530 -01 -0.117609 -0.107912 -0.4064440-01 -0.3417170-01
7 0.249920 -'.164041 -0.3424940-01 0.140023 -0.202014 - 0.216306
8 -0.234161 0.1062640-01 0.123467 3.8554570-01 -0.4914520-01 -0.2117720-01
9 -0.274914 0.7662840-01 -0.129753 -0.169748 0.348626 0.343368
10 0.163045 -0.304380 -0.3668910-01 - 3.3235950 -01 -0.543620 . -0.638947
11 -0.2565230 -01 0.4129860-01 -0.2852480-01 0.4069270-01 0.214082 0.386111
12 -0.277339 0.625842 -0.2536970-01 -3.261592 0.997729 0.978128
13 -0.291132 , 0.639370 -0.431766D-01 - 0.245333 0.972828 0.968315
14 0.8982680-01 -0.192318 0.111306 0.162705 -0.180981 -0.194505
15 0.505624 -0.312704 0.9579590-01 '0.705686 -0.349004 -0.354578
16 -0.209103 -0.356389 0.9547380-02 -0.2576040 -01 -0.191794 -0.200968
17 1.00000 4.333457 - 0.1697190 -01 0.505677 - 0.282676 -0.271618
18 1.00000 -0.5991410-01 -0.231132 0.627073 0.599708
19 1.03000 0.7731320-02 -0.3075210-01 -0.3429260-01
201.00000 -0.259979 0.237981
211.00000 0.983671
1.00000
0
378
Occupation Stratum 12-14 White Females
VARIAUE
SCHOOL
MEAN
11.426Cd6c6
STANDARJ (AVIATION
1.937899340
WAGL 2 2.354C56522 0.9197505134
UN.IJN 3 0.10434182E1 0.3070491430
t110<50 4 0.5211351304 0.5017133117
S. 16 5 0.20001.00000 0.4011505554
EXP 6 26.19130435 10.74006734
TRAIN 7 C.2C36956522 0.4081548591
PHY DM 8 1.4347E26C9 0.77380.34760
NEG WI 9 0.4173913043 0.7721308940
SVP 10 6.275652174 0.4610421651
%1TMOLC 11 0.139826C8700-01 0.25105036400-01
u&ncc 12 0.7033391304 0.3396325385
%13FOCC 13 0.20704347130-01 0.43677240030-01
CA/P 14 12.60935380 25.23131698
AV8MPF 15 0.29h4168116 0.2595031891
FCSL 16 0.27/J6618456U-01 0.57035768290-01
AlPAS 17 0.1676E6554.5D-01 0.18976085440-01
;;jr%i 18 0.43416c4462 0.2133400297
EDS16 19 2.2CC(C0000 4.569540650
WOW 20 0.51413913040-01 0.1801303133
UMOCC 21 0.73204347E3 0.3370630216
%MINUC 22 0.74602C0870 0.329 ).592867
396
C041-LAIICI% 97441).
vAotALI.Ati"61a
1 1.000007
3
4
5678
2
0.2426631.0 OUZO
3
-0.1666450-020.1701490-021.00000
4
0.4903320-01-0.3397730-01-0.1485440 -011.00000
-0.110416- 0.22J633-3.9955400-01-0.8703080-01
1.00600
6
-0.38361.70.3577880-010.110933
-0.3984060-01-0.2520090-01
1.00000
7
0.7512500-010.2712)93 -01
-0.1052960.6.332360-01
-0.9629110-01-0.131253
1.00000
-41.407302- 0.4034S80.6581210-01
-0.2455940-010.8464990-010.317106
-0.9539730-011.00000
v6RIA61.1hLm,AR
9 10 11 12 13 14 15 16
-(.411.tni -0.3'..5460 -;,.564695 ..47/53(. -0.155014 0.9113300-01 0.0601230-01 0.214081
-C.55t..66 -.).1C 10.1 -0.54315.9 0.316656 -0.445 /55 0.2296b1 0.4C1340 0.161049
r.1I(NII '3.151Gli 0.1521,h0-01 -G./C.3650 -0.7026:50 -01 0.105:32 C.259326 0.16062;33-01
4 L.4410300-01 -.3./132530-01 --0.41(i'.60-01 0.124407 0.4192/0J-01 0.4940260-01 -0.2/06460-01 -0.693762D-01
5 C.2C5257 0.4546420-01 0.116050' -0.105605 0.663080-01 -0.6923960-01 -0.132018 -0.776119D-01
6 0.242039 0.253569 0.271,542 -0.20i.839 0.1492o7 -0.217020 -0.185734 -0.5378190-01
7 -C.114055 -0.11.677 -0.191115 6.233428 -0.116411 0.1737550-01 0.4900620-01 0.4754790-01
A 0.4366124 0.7517:'4 0.095806 -0.539471 0.745056 -0.199511 -0.217771 --0.3785920 -03
9 1.00000 0.2b56(e 0.8121092 -0.565079 0./15079 -0.211550 -0.349416 - 0.8482060-01
1.L000O 0.4755720-v1 -0.650C17 - 0.228273 -0.210308 -0.9827300-02 -0.6612310-01
11 1.00000 -0.431006 0.904065 -0.195804 -0.306211 -0.2111620-01
121.0:,000 -0.122908 0.201566 0.8439200-01 0.376:110-01
131.00000 -0.137366 -0.329888 -0.5925270-01
141.00000 0.527038 -0.7520120-04
151.00000 0.6135510-01
161.00000
VAP1A0LE70,v6CR
17 18 19 20 21 22
C.1 :??4,1 -0.4( esan c.4476501-01 0.4184520-01 0.435067 0.406013
( ..4 :7' .) -0. 14',715 -0.11, 219 0.2640GuJul U.4634(.6 0.235315
C.13''.15 -0.,193.50-01 -C./752400-01 0.637099 -0.219310 -0.217001
4 C.371,570-01 0.11674 4 -0.0035e10-01 0.1100210-01 0.130764 0.130427
5 -C.77758.in-01 .1511"19 G.767112 -0.3611290-01 -0.9600910-01 -0.9122560-01
6 ..i(e..510-01 .4.1 flit -0.1.06360-01 0.664040,9-o2 -0...6046/ - 0.245070
7 C.91 .34290-01 -0 .1/.8135 -0.6!442100-01 -J.4676960-01 0./1d627 0.208845
-1:.394451 4..165410-01 -0.3426520-02 -0.h116/b - 0.352330
9 S i 4 I.) 0.' 1 1 IP.) 1:.1253' :3 -0.7343050-02 -0.57612 -0.425744
IC .) .4. 03141 C..104270-01 0.4666390-U1 -0.496137 -0.901597
11 -C.3 17544 r .o,Stit 3 C.3604550-01 -0.9243330-01 -0.517120 -0.247964
12 C.1 1 si..91 -0.350-7520-01 -0.7843450-01 0.991697 0.983246
13 -L.31C lob .30141.8 C.1533q13-01 -0.14070/ 0.5137010 -02 p.1464800-01
14 C.( ;sc.> I -0./0.12(.7 -0.5489...0-01 0.2694(7 0.245700 0.236153
15 C.3(145 1t, -0.145019 -0.0072670-11 0.425509 0.4728790-01 0.19904710-01
16 C.6064100-01 -0.215945 -0.6911433-31 0.0009010-01 0.3030040-01 0.2934610-01
17 1.00000 -0.381619 -0.4700473-01 0.209609 0.130393 0.104478
le 1.00(1.0 . 0.9(76420-01 -0.151638 -0.517100 -0.497639
19 1.0;..000 -0.7507943-02 -0.3396070-01 -0.3194930-01
7.31.00000 -0.94875331 -01 -0.103951
211.00000 0.997193
221.00000
73- 24
(V
.0-
380
Occupation Stratum 12-14 Cross Race-Sex
VARIABLE
SCHOOL 1
;4AGi 2
UNION 3
MIG<50 4
S 10 5
MEAN
10.342682933.233915610
0.38414634150.42560915610.25497510488
E.FANbARD DEVIATION
2.3499424821.175864652
0.48668962570.49473683460.-4 60586150
EXP 6 28.095121)5 11.46404798RACE 7 0.37904878090 -01 0.1908404674SEX 8 0.14878)4878 0,3560891571
TRAIN 9 3.3024311244 0.4595946639PUY DM 10 2.5)3658537 0.8970097008NEG wr 11 0.8731707317 0,9726477565
12 6.896f:17561 0,4544012312
IM:OCC 13 0.25736985379 -01 0.23041797309-01.A4POC: 14 0.1440500000 0.2730532581)5FOCC 15 0.85317C73170-02 0.27582137860 -01
DA/P4 10 33.82430662 73.'55092587
AVEML)1I 17 0.4275130408 0.3133135309FD 1q. 10 0.4Y48180212D-01 0.86266114170 -01
ATPEAS 19 0.43027926830-01 0.1833778735D-01
OININ 20 0.3079598231 0.1768430866EdS16 21 2.601219512 4.649444776
U'::CiP? 22 0.212265041 0.3323468487
AN:OC:: 23 0.152 5317013 0.2052737215
%MINOC 24 0.1733182927 0.2818013873RAXqEX 0.35365853661)-02 0.92054583150-01
398
.--i 1
oz) cs,
cl
00000't9t
L0-aus6L6'o- 00000"1St
tS9S11.0- S90Z00-0 00000%nt
10-019156Z-0- 06019E-C 8tfLEZ-0- 00000"1It
t0-0£606/00- 9SSS56'C- EtZ/99"C- 10-07P9fL6-0- 03Con't 7t
to-005169c0 et5ELt.0 ttS91Z-0- 00E9E1'0 btSSSt-C OnOK't tt
6E9SSt°0 t0-CP(7.£9V0 SOSCLS"0- LSt90.0-0 SIIS101-0 vSPL6t*0 COOOC't ta
03-009668E-0 016111'0- t0-07059t90- t0-CC0900/"0- t0-CP9t06S-C 10-1t96Pbt-0 to-090/07.9'0 OnCO't t.
LUtZ1"0- 0StL004.0 9ULf8"0 tS/Ltt°0- ISLSRS-0- Z/69S1-0- L6S9S6°6- tO-IsS7796°C- t.
10-09v79850- 0Z£90ZC 10-CIA5ISt°0 11.0661°0 t0-a6tta,t6N., 1,0-(166C446.0 tt97ct-0 IC-cftt64te- /
Z0-0090969°0- L0-3/1E079'0 t0-CIESIZ06"0- 9567c1-0 t13-a987099-0 t0-nct0t0S.0 t0-e0Stc.Fv t;?ttizzon- 9
t0-ccren9ro0 t0-actrs9co t0-USZthSti*O- 519111"0 t0-ctSSf.C7"e- 76061t.0 t0-19t0FL9°C, t0-COE(PLE°0- S
ZO-COLSLZVO- tO-CO9f0LST 10-C969Zi6-0 10-Ct09,,WO t0-04bEi.t9-C- t0-C,.ttcL6-0- te-Gt0Sof'f- ProtrtC- 0
St6PEt"0 9tEtLt"0- LSCREZ-0- P/C7.et°0 07,i-7°1°0 ct000t-0- .171St-f°( TO-CIZEZ'Co'f- f
66atlit'0 ERESZUC- ttat9Z°0- 95tOZ2°0- r9t6L7oc to-crcimsoo to-ncs-407ro rtir,vc z
to-Qom/coo° f111 61'0- tb/SZt°0 96fItSE°0- 10-GGLttLtC r_Stc4:1'0- (S.:JP/VD- 691fSt'c t
e=e6fili
9t St tt ft Zt 11 01 6 21cY'rvik
00000'1P
t0-08z06zh0 00000%L
1.0-Crt0009"0- Z0-OtS609S.0 COCOO't9
10-e66a69t°0- 909Z9t°0 10-000/6Zt "0 00003"1 S
10-09Z9/9/.0 t0-06tLE8Z"0- t0-Gt6VitS.0 10-nS899C5"0- 00000% t
t69StZ"0- 1.0-C9fL9LL'O- t0-01£1912-0 SL6tit.0- Z0-09/t0Q6-0 clonoirt r
ttaSSE°0- tR9PSt"0- t0-OPECS6t"0- fi69fUZ-0- ttSt9tC- L690ft"0 0000C't 7.
10-COL690/.0 ChZ601-0- ttS9E6'C- tZ,W.t.-0- t0-ORLILSCC- 10-J0LRUS'O- 11509r0 000C2't t
npt:rt8 1 9 S ti f Z l SIEllOtA
x1 E111. C01.1V11.1P3
t"4CO
4-0
0U
00000'1 -
ELL966'0 00000'1ZZL6t7Z*0- 0LA6AZ'0- 00000't
10-CZtE00ZO- t0-0A960AZ*0- t0-06SLLZS0- COCOCt
StAOLZ'O859901'0-0t9ZWO-tA9CEZ'O-S7011Zt'O-fES6tS'0CO6606'0EALOW0-C80669'0-ZZ9t01°0-tZtEtS°0-
tALL9ZO 9,11671'04 10-CZOCLOO66SL01'0- ZUZZZO 665E11'0-6970W0- 10-0856011'0 ZO-OL5ZE61-0-96ZLWO- ZEAtZ9'0 t0-0Z6LZW0-tLtOZE0- 9ZSLtC0 to-OC9fIt6A*0
tt6tRA0. 90SZ91'0- t0-019191E'0SL0966'0 OZZAAZ0- t0-CFP9C9Z0-99106I.0- t0-0Z6tSACe 10-0t8ArCS090089'0- t0-OSLC060- ZO-040CL90-AZA0610- t0-00ALL06'0- t0-crtontfoIZSEA6.0- 6871(6'0 t0-01.10SrZ0
00000'1P69ALt'rSCO9W0
t0-OtS3C65'6-SCt?(I'C?EtZit0ZC9611.'u-
Eu(91.'0-10-0169;01"C-
erfisr-c
L0S0(7'0rfr4Pt't.-
SSSL01'0IC-rOSL(1,_0tf-cnr6cto.c_
prtt;?'0Sit80t'0
t0-c97t71Z0917ezt'c
10-CL1PSsS-Ct0 -C[687106-
StC4S1-0-t0-r?Pci.1%)-
SZ
SC.67
tZCCt?OZ61EtLl9tSnt
itZttt
Ot6
L
:11r V 1 gVA
001.00%LISLWO- 00000'1c975ZC- 81,091t0- u9000'16LOSE1.0- SC61a'0 t0-OZCL6AS0 00000'1
9SSL6tt- 0grat0- 70-0091801J Stat6o,0tzoc610 940z67.0- t0-aubst,60- At9ASZ0-
75E6C7'0 CASSAt0- 6cillOt0- 9h7LC7'0-
ZO-05COA9C- ZuNLOZO- ta-UCICeZto- PoLLQ-Z91L61.0- A0A1i.ZO 10-001LtrAe te-CuFCL6A'0-
to-ast ci Lr 'c oqz,.Ptzo- 7o-aisckni)- t(-CrSALZUC-Et L8Ct '0- t0-0Z8LL9t '0- ?At( (1 '0 Lt.9L')Z '0
t0-03A07.A9L'0- L0-006(9600- t0-0t685c.S0 10-0167971'0- to-0z4sbcz'r- 1C-0n6r1rZ0 t0-OCS91/10- PECSIt'C
Le6(0'0 ttItAP0 50700Z0 - CO-M7(00- 0o7L 7( "0 77S1,9t0- tu-n?(Asto- SSIC6t'°-
t0-09t6tSS'0 t0-0A60L010 t0-Cfm/Ait'0- Ct6Cit'0 9(L77tC 10-06161L0'0- ZO-C87Lt16''; to-atepf.L4'u-
t0-at.zzt9C0- I0-ontothuo- t0-a8 St0It'0- t0-CN1f6A0- t0-CrociAtC- 10-OtilCo,.:U to-Ott0:C.:e to-rfrItn-
t0-0AOSLSZ0- L0-091909E-0- 10-00R6SA9'0- 1,91.L66*0 tO-Ict9Cc,h0 7c(;At'0- Zo-C(AtA01- IC-CSAL916'0-
10-00E16;6'0 10-O9tfOC6'0 10-cA0(6660- t0-00(t6P90- 13-Co617SC LO-gto7rOf°- t0-01r0iiti- t0-0SZAtu8"(7-
9LPSPZ'O- t8ZZ6C0- LSZ60t 0 95t6t t '0- LI gt (A 'C- C6t1140 tO -rin. 9 tAF'( CSSStr 'e
t ASSOC '0- IP( APC 'o- 1S8SE t 'e c.OF Act '0- (4'7' , Li 'C- Lo tf,4,7'1, it .; Ll't, CI. It.t ' r
t0-C9ZS6LL '0 68SSOt '0 tO-of t (.98s '0- t0-Ont i Opt '0 10-Oro. ARPA '0- tOuvtu 70.-CA1t6'0- ti-q.bitts(%
AZ CZ Z? t Z OZ 61
67.
ZZl4C'61atLt91StAt
Crtt01
L
S
42;f1.1r9
$f Lt 37rY:-:YA
383
Occupational Stratum 15-17 White Males
VARIABLE
SCHOOL .1
MEAN.
14.21254355
STANDARD DEVIATION.
2.656956919WAGE 2 4.831254355 2.038005974UNION 3 0.1114982578 0.3152979263MIG<50 4 0.2682926829 0.4438647069S. 16 5 0.1846689855 0.3887065855EMI 6 22.43902439 10.60325239TRAIN 7 0.4494773519 0.4983098062FIN OM 8 1.5365E5366 0.4995307366NEG WT 9 0.13937282230 -01 0.1666646362SVP 10 7.477700348 0.4333851233tomocc 11 0.93170731710-02 0.75043452080-02T.WFOCC 12 0.96745544600-01 0.1127932453TBFOCC 13 0.28710801390-02 0.84556080540-02DA/PW 14 50.22205967 108.2667440AVEMPF 15 0.5541205575 0.2794237727FDSL 16 0.45198488C70 -01 0.731/900541D-01ATPFAS 17 .0.47249128920-01 0.18929254630-01VIININ 18 0.2821773328 0.1419488676Ens 16 19 2.666411150 5.743217312UNXMPF 20 0.64912543550-01 0.2064812091TEMOCC 21 0.99616.7'4740 -01 0.1184538186%MINOC 22 0.1089337979 0.1210161421
..
401
CORkELATIC4 4ATRIX
vAtatA0LE
12
34
56
78
NU
mbE
R
11.00000
0.433729
-0.224386
-0.6330120-01
0.599980D-01 -0.290918
-0.8554820-01 -0.2035330-01
21.00000
-0.163786
-0.137325
-0.4707170-02
0.4498990-01 -0.7960690-C1 -0.136739
31.03000
0.110300
-0.61C0330-01
0.6688410-01
0.1312430-01
0.107209
41.00000
-0.105782
-0.5260500-01 -0.2544860-01
0.2654033 -01
S1.00000
0.9104200-02 -0.3289520-01
0.4611640-01
61.00000
-0.189681
-0.1961810-01
71.00000
0.5310310-01
81.00000
VAR1481E
910
11
12
13
15
15
16
N048E8
10.5641240-01
3.357VC8
-0.115162
-0.220490
-0.4512110-01
0.27/9450 -01
0.8641470-01 -0.4309770-01
20.4050680-31
3.289109
-0.205257
-0.215080
-0.104085
0.16984,0-01
0.117782
0./932270-02
3-0.2967553-01 -3.5082860-01
0.120959
0.6254350-01
0.1590250-01
0.4761550-01
0.34818/0-01 -0.2900470-01
ItiPh
4-C.53/25/0-01
3.1848760-01
0.43/0150-01 -0.1888043-01 -0.513J95U-01
0.4106460-02
0.4249940-01
0.590518U-02
95
0.5807623-01 -3.2320740-01
0.5057830-01
0.150313J-01 -0.25/0950-01
0.20040
0.3105600-01 -0.2628600-01
6-0.5887450-01 -0.18/1040-01 -0.1315633-01
0.114797
-0.2003530-01
0.3436213-01
0.4353003-01 -0.2001150-01
7-0.7569360-01 -0.1299C2
0.7021/90-01 -0.10687/
-0.5392620-01 -0.6796930-02
0.125160;
0.5166853-01
80.7784990-01 -0.228792
0.438,45
-0.125090
0.114943
-0.6407800-01
0.2587801 -01 -0.1597790-01
91.00000
0.101134
-0.7064050-01 -0.7048990-01 -0.2849180-01 -0.2065570-01
-0.3309233-01 -0.5103080-01
10
1.00000
- 0.391196
-0.186984
-0.152783
0.159791
0.7441083 -01
0.2742260-01
11
1.00000
0.303501
0.346529
- 0.6740370 -01 -0.8175000-01 -0.6839390-01
12
1.00000
0.648161
-0.5568010-02 -0.119512
-0.5684820-01
13
1.00000
-0.2909530-01 -0.149540-01 -0.5711630-01
14
1.00000
0.445996'
0.6626760-01
15
1.00000
-0.2166510-01
1.00000
VARIA8LF
17
18
19
20
21
22
N04318
10.8631130-01
0.1968750-01
0.175272
- 0.173357
-0.213174
-0.215810
70.160617
-0.7501250-01
0.1924460-01
-0.133110
-0.212232
-0.220463
30.15)1820-01
-0.4252470-01
-0.9940710-01
0.888099
0.6068990-C1
0.0690460-01
4-C.190819
0.3434600-01
-0.919270-01
0.105283
-0.2164650-01
-0.1815540-01
5-0.8198660-01
0.2969220-02
0.984556
-0.3650250-01
0.1241180-01
0.1534970-01
6- 0.129963
0.4/27710 -01
-0.1102320-01
0.5850420-01
0.107831
0.104766
7C.2641793-01
-3.119888
-0.1854190-01
0.3427290-01
-0.131309
-0.100681
8-0.9427553-01
0.114186
0.1)35550-01
0.153858
-0.1/0913.
-0.8116830-01
9-0.7,640^10-01
-0.9358000-01
0.7701530-01
-0.2033140-01
-0.:,415530-01
-0.7207040 -01
10
-0.1163470-01
-3.1(4980
0.1853300-01
-0.6213303-01
-0.183954
-0.209209
11
-0.6471910-01
0.201534
0.5017963-01
0.7442690-01
0.313734
0.369097
12
-0.3884180-01
3.131022
-0.670120-03
-0.4200900-01
0.598523
0.996185
130.7204340-01
3.109916
-0.3121920-01
-0.112131a-01
0.689142
0.696028
14
-0.232797
-0.26088
0.197531
0.130150
-0.7318350-02
-0.1140220-01
15
-C./844520-01
-0.226171
0.7849130-0(
0.165452
-0.114868
-0.117504
16
-0.140003
-0.240545
-0.1555610-01
-0.2510863-01
-0.5820970-01
-0.6121040-01
17
L.00000
-3.175142
-0.8216560-01
0.8600960-02
-0.3184100-01
-0.3518090-01
18
1.00000
-0.2775160-02
- 0.9376420 -01
0.112607
0.142790
19
1.00000
-0.5705070-01
-0.2814020-02
0.3523560-03
20
1.00000
-0.4080200-01
-0.3501210-01
21
1.00000
0.998265
22
1.00000
385
Occupation Stratum 15-17 White Females
A r
c." 4-Irtt
WV;F1,M1'1%)
4!",eqlC. 14r-vt,
T4A11iv,
7
1
4
6
7
0
1FAN
12.966666672.831000000
0.10000000000.33333313330.16666666672.5.71311333
0.211000e00001.466666667
STAN0AD DEVIATION
2.2815105431.063703193
0.30512157660.47946330150.37904902189.818818432
0.40681810220.5074162634
Mr^. wr 0.010 7.513133313 0.4256715326
1!e: 0.90313333330-02 0.82314926510-0217 0.3241666667 0.2454937998
.vorn7r 13 0.1106666667D-01 0.17101085530-01r1 / n 4 14 .24.22796968 32.88062392Av-.4nr IR 0.4112411111 0.2803001130
r.r.r1 16 0.28406134540-01 0.41191642730-01ATnrAs 17 0.48143333310 -01 0.18323143210-01Y111 %!! 14 0.1014732116 0.1359390593
10 2.200000010 5.067815962ivoyqpc 0.79844444440-01 0.2499296122
71 0.3353333133 0.2611869899TMtmlr 0.3443666(467 0.2597110441
403
38-6
Occupation Stratum 1C-17 White Females
4 .4 nSO 0 0
1
O C3 CD.3 4. .3 .1 7 - 0
m ) 1-en 033 7 4. .3 4. Ws CO JO /-0 r. - N 1.1
01 r, Int 41.4 fy md ns
0000000
Ct 00, P. 0l0.0 0 0 0
r, 0 .0 P. ^4 CP C)CY 04 4 r4
C) C) 0 C) 0 0I I 1 I
N . nC) C) 0 C) C) CI 0
IIIC) COO 0 0 C)
7 4 13 13 7 -. en en 4. n4 "0 PI .4 0 0
C) ^4 CD r. 'no r- tn 41 C)
4. 4) J.r on Pn n- .4 on y. .2 C. .4 .3 0m 41 el -. nD ry 0 .- CP .4 II
0 .o004100..0.411, a .40000000000000001
.11 . 00 oI I
e- C> -J .0 7 ->O r4 rn C) 4 .4 CIO m 0 4. Cl .3 7 m 4, 4 U1 C)Cy 4, CO CJ CP 2. W, CD C) C1 a. C)p, ys f 0 .4 el ry 1. 0 .* 4. 0yrs 4, ,c) C) cy 41 .or S uft cc UN 0-4 r4 0 0 r4 el 4 4).. .4O CD C) C) C) C) C) 0 0 C) C) 0 0
.4 w48. 04 04 .4
0 0 0 0 0 0 0I 1 1 1 1 1 1
0 0 C) n C3 nP. c 4. 1 CO CP 14 C. 22 0 C) 0 0 P1 0' 4) .., eq 0 ,, ..., us r .4 .4 .4 w4 4) P. rj On 4 ../. on N P. UN .0 N04 , 4. to e n m 0 P. . ...- .1. P ' 0 = es s; ...1 4. 0 Cr r 1/ P. C) .1. sr 0 .. 4^ a 0 ... P4 C. in .4 0
.. .4) ..,, N 4 3 u s 0 ..r r 4 T A ) ni - e 0 C C .. 0 4 ) r - 0 e 4 0 .0 C. .0 V. 4- an 0 C. a .3 14 Ca 0 ...4 1.4 C el UN UN C)
.. r ) .4 el 0 C) ... 7 n4 13 rw - 4 41 c , ..? 0. .{.. C C 4. 0 PNI on ' C O ' ' 0 41 7) CY 0 1. 041 C) C) 4 CP 0
.... 4, C, 4N 0 ...) t- ,, . e - 4 4 4 r4 ty Co 0 02 a' P. x ...r J-. a) ....1 0 (1. ,... Z! MC -4 en r- .r POen.. .4 .4) ri
T- ... .4 ey CY n+1 CD 4 4) .. N 7 P.. w4 ry ... .4 oy 4, 0n -. C C ,. 0, ..- o, ,.. .. ,.. C'
. . ..4 O .... 04
0 0000 0 0 0 0 0 0 0 0 0 0 0 0 0 0000000000000000000001 1 I 1
1 1 I I 1 1 1 1 I
... ..4 ...4 .... .. .. ... ... .. ..4
0 0 C 0 0 0 0 0 0 C
i i I i i i i I i I
0 0 C 0 0 C) 0 C5 or s n4 .4 .4 C I nl .y ry en 41 C . C Z en 0 C r 0 1 0 NS 41 ga Cr ar P 0 NO .0 C) -. C .0 ../. .3 C) oo JOt. P. f es 0 r, w, 1) 4) 11, st 0. CY CC 1.. P Ca
e. ,e 0 sp ,f 11 0 .0, Us ,r 4. e. r- 0 011.1, ... 0 0
nr -. 71 c..... ../. o . ..0 0 ...c ..? ?...e 0re.4 ...r es, .40 . on .4 V. 44. lr T, 4) 4. CD ,.1 .3 co en. .4. ...2 I- r. en CO 0
w n 0 C P 4 ; 0 4. C) - 0 -. .3. .3 ...rry .4 CSO C) CY er 4 C. 4 4 c 4 -, 4 ',0 rj -... . . .0 I 1 - k t s .4; 0
4, .1. 4- C1 0 .. ..: .0. CC 4) n .0 0 1.- .. 0 1. e- r. .... PJ P. -. .0 0 CP C .... NT C. '., C.l rsi IA 4 0
.1. c .4 .4 ce ... -+ rn .. ..4 ev 0 C) CI .3 Cr3. 43 ... r.... .... t. m Q v. .... C. C. .... en in. en .4.4 .4
. . ..4.4
C 0 0 0 0 C 0 C 0 0 0 0 0 C) 0 0 0 0 0 C Ca 0 0 Ca C) Ca Ca C) C) Ca C) C) C) C) Ca C)
i i II I 1 '
I I i 1 1 I
4:1..,, ... ... .4 o4 .4
..4 00 0 0 0 0 0 0I
1 I I I I i I i
CDre m c, ,:, f., 4 r C . ..,
J cc r. 0 7 ,0 .. na to t- cC f- .3 0 C) cr en r- en e- le 11 es r, 4 4 c1 CP Ct ..r 1, 1.- CD
43 23 C' 0 ,r 45 4. C) e. 03 us 4. .. 4, 0 .. '- .0 0 ..4 .2 0 V. 4 ce CD 0 4. r. .3 rs 1. 4) C>
,r 0.3 .on 0 ry r. .- 40 r4 0. or ev r) r, 4 C) C) C., e- tr en 7 ..r 4. ,.... 1- -. 7 e- C .3 C -. .r0... en ...r 0 .. 4) I' x - -. f. CC.. Ltn --. 0 INI CC 4. Iv 0 2n a*, sr 1-2 u- en 0 .2 .1 a c o4' 4 CP 0 r4 c N 4. I. i "r 0Y+ 0. ...... 0 4. e "- 0* .: .: .1" 0 7 ../..... - Cr 0 0 Ic 1.- 4 CCV Psi eV
0 ..::: .4 I", .4 4.4 41 0 .. N .. .4 7 L. .3 .4 .O ..4 .4 .4 v. ... .. 41 N w4 .4
.4000000041e.. .. OoOOOOO. 41*. .4
CD el 0 C) 0 Cl 0 0 0 C) 0 c) T.0 4-1 C) C. C C C: C, C, C. C C CI 0 0 C) 0 C: 0
I II 1 I 1
I 1 1 1 1 I I I
p.... ... ... ... ... 04 G.
0 0 0 0 0 0 01
1 1 I 1 1 1 1
o a n a a o a ac t. 0 It 7 e. 7. 0. tv en C 0. 0 r .. 7. a: .. f- 0 u- cr C ee u- m C.' 11` a a4 r. o r, 0 4 V' e C. es 00 C.0 C' C f- 7 C "*. .1" r 4 of a en e- ...1 a- 0. 0- 0
e. 4 C .4 .4 ... C. .4 0 .1. C C) CP C a ..., P. ... 0.. 0° C. ".". .1. r- -- .. n. en eno 0
M .. 0 .4 C 7 4 ev C)... v. .-. 1.. ... P.. N c t ... P.- Cs N.. ... Is (1. 0 .0 0
.r 1.- 0 nn ... 7. en .. .3 m ,r -. 0 r- .C. .1 we a* .0 es. r. .0 .1 kr -. - CC - 0: kis 0.... ... .... r "" 0' .0 Y- r 0 4' -.
CO.4..
..
CO CCCCCCCCCO CCCCCOOCC OCCCOCCOO1
1 1 I 1 1 1 1 I I I I
.4 .4 .4 .... e., .4 IV a.. 04
C 0 0 0 0 0 n0 oI 1
i 1 1 I I 1 I
Mo ...-..c C C cc 0
CC rs C r sr V S 0. en c 0 0 r n .- C.1 - 4. r n cC C.. .0 0° 0... NV 4 - 0
st, C C 4 - C C 4 0 - 0 u- e. ..0 C n. - C te C - ..: .1 N.: .410 0N 7 C ...et's.... rl ". 0 a' F .. ..: 1.,7 .1:-v c' :'.: P: ;:: fr ..7. f-::- 'C :!: j. E:
.-. 0 C 0.- C C ... tt 04 a' C. -.0 C. -. c ,:, .. en .. cc ..1 0 C -- tr .- ..4 f. 4. C 1\ ,... ". r. -.. r II 4, C)
V.(... .. ... ..,C ... .- fr 11, 0 . e c.... .r -- Cr .04 .4 r ,.. o .... v. .... er,
... . . . . . . -. ............ ....C rereCr,reC e c` r r r e. e. r. r C. r r corer
I 1 1 1 I I 1 I 1111 II... rn ..., .. .-4- r C .- C.
I o 1 1 1
C C. C C C.,r.- .- o cc 7 r e. r. C u- u- C 7 r- 0. x . .
p c. . C . o e .e e r. I V r . U 4 ,r 1( - c CC r)
Vc e 0 I. V t .4 P. 4. 4 c tc-ccor-a
41 C.... P. P. C ...0 C C 0. f. rCCCee t 0
a r c c C C 4. .. - 0 0 k? .- 0 r Cr r r r r. r r. C C -r r r r rver..)rr4r,r
: ... .-
C cc,..c.cctcc C.CCLC-Ce CCCCCC.C.CC1 1 i I
...
... 6 4 6a .. .0 -cr-1 .. . c n .. - r,, .. 'I .. c a r - e - . v . 1. a C r .- e e ,f4 r . ,11..P.c c - e. . 4 Y. r a reef. a. ... :
gm ..... .. r.
... 7 . . >
404
vAQTAPtF
387
Occupation Stratum 15-17 Cross Race-Sex
MEAN STANDARD DEVIATION
SCHnOL 1 14.09006211 2.639496435
WAilF .7 4.628322981 2.043987532
UNinm 3 0.1149068323 0.3194058308
MIr<50 4 0.2763975155 0.4479118082
S 16 5 0.1801242236 0.3848890079FYI, 6 22.71118012 10.51451152
QACF 7 0.15527950310-01 0.1238323690
SrY A 0.c9378881990-01 0.2996356863
TnA1M 9 0.4254658385 0.4951829648owl nm 10 1.531055901 0.4998113071
NFr: WT 11 0.12422360250-01 0.1573763725
SVP 12 7.476260870 0.4322217690
T.P"nCC 13 0.93229813660-02 0.75398029870 -02
crvFnrC 14 0.1199813665 0.1461922292
714-cCC 1' 0.3613540373D-02 0.97812137570-02
FIA/TP4 15 47.30596888 103.0200964
AVF:inF 17 0.5407351967 0.7819909417FOCI, 18 .0.41449043370-01 0.7050367285D-01
ATPFAS 10 0.47670807450-01 0.18512469960-01?up1IN 20 (1.28355C2783 0.1404519636
Frci 71 7.599379882 5.642329968
iroXYQF 77 0.6964068323D-01 0.2153510186
Tr"nrC 23 0.12261490f8 0.1515361206
7tArl' C 24 0.1319378882 0.1550108102
RAxSPX 25 0.62111801240-02 0.7868818624D-01
405
388
Occupation Stratum 15-17 Cross Race-Sex
A A A .400000 0
P. N 7 ca POM 1 fa f caCO M PA 01 el ly .0
.4 A .4) 4 1 .0 0-r 7 0 0 0 A N 0- kr) n1 9 Aite,e4o0 0 0 0 0 0 0
I I
.4 4 A 4 .400 00011 111C3 C3 0000' rn PJ JOey 0. P. C) .0 C)
P- r4 to or 44 .o CDO A y P- C) 0 001rn....., CO 0001C) CD 0 CD CD 0I I I 1
fy .4 A AC) C3 C)11110 CD 00
*N1 .0 .1:2O 1, C3 JP 0- 1.nI 0 0 0
ci .r0ftf0 .40 0 0 0 0
I I
NA000I 1
0 0V CD o% C)
01 0. 01 d 0O U oN C 0 0
0 A N C4 .4 C .4.4
0 0 0I
OO co P 0.OT 0 0. 04) Pro% C)CT V% O C)ec .4 4.
.4000I
4 CC0c 04 VA CP-Ctf 0
ry w40 .40 0
I
.4 .4 ey A R A Pv .4 A A o0000 C) C) CD C) 0001111 1111( 1 11 10000 0 0 0 0 CD 0 C3 0 C30" 4) 4 n't .4. J OJ oy o cp C) nn 0P- CD kn .4 en en kn 4. 4, el .4 .0 V. 03 01 C3
4 O e. rn 4 0 Tr. A yr g) ry 41. 4) 0MC A Cr. 44 a. .1 4 f.1J ry TNNNO(7. .4 4 P. OP 41 P. 4) .0 MJ orrn J yr ry ry .4 en P. NJ A A V4 1.1
4.0 A
C) 0 0 0 C) 0 C) 0 0 C) 0 C) 0 0 011 11 111
.4 ey .411J A 1,1CD CD CD CD C3 C31 11111O 0 0 CD CD 0un CD P. CT .4 4p .4 4) 41. cp p Of A0
NN yl p. .4 U1 .5 .4 on f4 C) NOM crrnen ACM A A 4100. -1 0J1 010
M4 . p, pm 55 OM J0 ,r 4) 43 C341 01 .r o4 0. 0 1 CD 4 CP CD P. rn 0MAN fY y1 A fn fY .4.4 N ry ey p A0000 0 0 000000 00
1 1 1 1 1 1
4 P./ 4 .4 e4 eo0 0 0 o 0 01 1 1 1 1 I0 0 0 0 0 00 d0 4 P N0 N 0, 7 17 N 0
4. P. ey pft CD r4 .I. cm Pen 03 0 on 04 .. AN an .0 4 P4 rm .0 CD an .4 ei 004 P.. Cr. P. 7 ry 4) c3 r, 4. 44 on Cr 74 C37- NM C) .7 P. 01 P. 4) 01 al d0
A PY A P. f4 .4 ff) 4 A 41 4 AA .mef00000f0f00C C ) D C D C D D D D D D
A1 1 111/A A A A A MI A0 0 0 0 C 0 0 0I 111110 0
0 0 0 0 0 0 0 07 rN , ry 04 A .0 Pm Pm Y1 ,r 0 0.0 c", o" C. 7 P- C 0 Nn . J0
0.% of C. 0 4' - 0 r -.....r o -. o... 1 4 .0 AN CP 0
4 ) P r A J PO 4 1 . 4 4 73 c0 41 y5 CP .77 AA 41 CO 4 A P. ...1 4r 4) en
ow.o 0 o 0 o 0 o 0 o o 0 o
I I
0 0 0 0 0 0 0I 1 1 I 1 I .0000(3C) Cc) y 4.1 A 4) 4) CO 0 CP C)
C r+rA rnNN P. 0 4. 4. ct) 0N. 4 09 NNP0 P 1. 0
C 0- O. co *2 u3 O 00 4 0NM CD .4 CS y) A en CD 0 411 0
f, ANN u. N P4 CP
0 0 0 0 0 0 0 0 0 0 0I 1 I I
.AO C 0 0 0 0 0 0 0 01111111111O 007000000
LP A 4 4 r. o. N 0ce. ey o
C- 4) CD 4) CD ,C PM CP oNC 4 ca c3 0 NN ON 04.4 z, en 1, C Ce rbO7u JN 4 .0 4 7 e. e.C) C C) C) 0 0 C) o 0 C)
11 011144C 7000001 1 1 1 1 1 1
C COOC)00JO ..r CP ea cr 4m 4 CP 01 MJ C)
CO 00 or .1 PY CO 0 CT, 01 4) C)
ts, N 0 C cr cC 7 1, 0 .01.. 0 .... .4 ,,, C. cc 0, if, 0/ JO4 0 v.: C. e. cC .. rn p- 0 01C.4' N. -.. - 0 1111 1.. ... N ...,. A 0 410.00f00cree-0c+oe-e-
I s
.. -. N ... -. Nr C r C. C etIIIII 1
M C C. C C C... C .0 fs. ct 0) 0. a 0 ocr e- ...,, ,:. c c .0 o P. P cl... A C *7 0 c .. 4 4 4 ct Ce C
C 0 .1. t. r. e. C .0 p. C2 C. 4 r c c r, o 4 o, " v , 4 N 4 -7 .... ilgfee0f0 ACm
CCCCO 0 O CA 1 1 1 1 1 1 1
om U. tQ -.. 0 .J CI... C Ly. 01 5 ... r. r . 4 y 4 . az <6 ... n r 4 it 4 e. a 0 c - ev f' 4 V 4.7 ... ''.
m. SCc .7 a Z'C
a< 7 <4- 7 7
406
$;11/
414
IIA
RL
FN
1148
FP
4 7 P
10
11
12
11
14 lc
1,,
17
10 19
1P
71 27
2-A
74
WigtARts
olgunrg
11
(1.1c/S4c
.*AN-r*01-01
!`'27[.19 -01
0.4121710-01
0.12,.0410-011
-0.11"",c10-11
t.114744
0a16A10-01
-9.7757400-91
A.15c4A70 -.11
-0.4117370 -01
- '1.106316
9.7795140-01
0.444ns
1.90090
's
In
-,.2071540-01
).4271150-01
-1.154(!910-M
-1.3062093 -07
-0.2144:40-01
-1.1117240-01
-1.1:124081-01
-n.711co00-01
0.1-0710,11-01
0.1717810-02
-0.4979480-91
0.2099390-01
-9.50.30340 -01
-0.6912219-01
-1.4410700-01
9.6862840-01
-0.9156410-03
1.00000
19
0.7333120-t61
0.145.,73
0.7;16402D-01
-0.P475100-01
-9.1 (1009
-0.11217e40-0!
0.116658
0.7607190-01
0.7919140-02
-0.6osiAlo-0!
-0.2682190-01
-0.5318100-01
-0.6097640-01
0.4015450 -01
0.4876410-01
-0.272184
0.1232600-01
-0.117962
1.00000
20
0.114417,3-01
-0.014n600-01
-0.2516,160-01
0.7060 060-01
-0.1014110-01
0.%06,,010-01
-0.251(000-01
0.4100769-01
-0.121!79
0.112247
-0.96660-01
-0.9138983-01
0.1331571
0.8635340-01
0.7597600-01
-0.261290
-0.202611
-0.234239
-0.192116
1.00000
21
0.125d45
0.2',09470-01
-0.105150
,-
-0.10071
0.904401
-0.15930-01
-0.5194860-01
-0.1165790-01
- 0.4554770 -01
0.469,510 -01
0.7574810-01
0.7914180-02
0.8123550 -01
0.6464180 -02
0.5490470-03
0.209211
0.1795800-01
-0.2964040-01
-0.104993
-0.1432520-01
1.00000
22
-0.1844S9
-0.113316
0.898904
0.9676290-01
-0.5399790-01
0.5095470-01
0.122761
0.3318020-01
0.1057310-01
0.160413
-0.2560540-01
-0.9321260 -01
0.8751840-01
-0.4957110-01
-0.2458400-01
0.11716/
0.22175$
-0.8830910-02
0.5307570-01
-0.6417480-01
-0.7073890-01
1.00000
23
-0.191309
-0.2649e5
0.1164030-01
0.5810700-02
0.10.1982'339-01
0.140104
0.3587120-01
0.427317
-0.172965
-0.4462060-01
-0.6220220-01
-0.193e01
0.171973
0.99q070
0.764713
-0.3237900-0r
-0.100269
-0.6802920-01
0.4452570-01
0.8706310-01
0.6189970-02
-0.4876620-01
1.00000
24
-0.194173
- 0.210913
0.1769130 -01
0.8593550 -02
0.2352630-01
0.116460
0.3721540-01
0.423641
-0.168002
-0.2209020-01
-0.6484030-01
-0.213718
0.21(977
0.997457
0.770799
-0.3422230-01
-0.101118
-0.7064550-01
0.4113620-01
0.9506660-01
0.1017970-01
-0.4404530-01
0.998851
1.00000
I-n.176q980-11
7-0.740'4110-91
.9564o711-11
40.1452760-Q1
-n.17(1654E1-01
61.3171750 -01
7c1.cl94R4
0.717943
0.11",P30 -31
In
0.7624 n,0 -11
11
-9.C75 ^9 m1-17
12
-n.47,4419-11
1114
_0.1A4p.,,541 -:/i
14
..(1.141,11,,2n-37
II,
-0.7403170 -01
17
-0.4S46c911-11
IR
-o.,,N6p^541 -nl
14
41.?306290 -01
7n
0.1 22 16141 -)1
71
-n.1647760-11
2?
r.7007070-11
-0.1819910-91
24
-9.1427050-91
2%
1.09900
cn 0 In 73 w X
390
All Occupation Strata White Males
VAPAAULT:
1
MFAN
10.94864'165
SrANDAn oiNIATION
3.0 34119469WAGE 1.42 3854054 to.? s 14') I
ON 9EM 3 (1.,,186486486 9.491717'1177
'I G< 4 0.4061081081 ..r9 16 1: 140
S.AT1h 5 0.?114524374 . 445331564 ')
EX0 S 17.74176173 17.15366031TRAIN 7 O.Y5/Z91273 3. 4372(sz79 4
PITY OM ?..36P1a6108 9403$3510925
WT 0.-;4541154054 1.904494530110 3.'61351351 02.5699,492
Svp It 5.,/6,q/595 1 . 995143903Tilly:CC 12 0.4.1=)342142.20-01 ). 52973345160~01r:10:1CC 13 1.12725021o3 1.1,483h4it105ne.:1F0CC 14 r.112/02C0903D-)1 '1.:'17543.311U-01DA/PW 15 14.9.36C7,90 :10.";56?7515
NYL/PW 16 35)753 11[,5') 5. 4 )6 )3480AVf7"PF 17 0.4:'1.-%10702/ 0,:qq78%/4472
11 W.1;1462196!)-(11 0.:.7)041'..9310-01
FrISIA:x 19 C./2217117410-ot n.q?)to114??0-01ATPEAS 21 r./.,.hv.21A2M-)1 1.1755434:50n-01MININO 21 0.2ri/775.62 0.15M1/5216EnSib 22 2.7/045)45 4.=055066517
HNXvPF 23 0.'00,15;1792'1 1.19717509211-mn0C 24 ^.17149C2703 ).19'6212790
PIN1CC 75 0.22/0264805 0.1971R35872G9 +SV 26 3.P27945946 2.7156U042GOXSV 27 2J,7L6C4865 11.00030345
408
C.1;
CP*Z4r1ATICk 4ATRIX
W.,41411 E 1 2 3 4 5 6 7 Cl
1.10130 1.412129 -0.218874 -0.135591 -0.150448 -0.474197 0.225483 - 0.333458
1.00000 -0.4900770-01 -0.153548 -0.112269 -0.111559 0.106515 -0.2;0593
1.00000 0.103967 -0.106101 0.641,5140-01 -0.8913650-01 0.333232
4 1.00000 -0.6345390-01 0.5991220-31 -0.83817/J-01 0.7006500-01
5 1.00000 0.2903040-01 -0.5465310-01 0.8508790-01
6 1.00000 -0.418:014 0.104261
7 1.00000 -0.6443150-01
81.001100
WV,IARLFr,"r6
3
4
5
73
In111211141516
-('.1.q/1-p.47cvl?9-11r.Q/14471-)10.371'171-020.9145739-010.40P8440 -31
-C.1724770-010.4891681.00000
10
).4-%-(015
).11t.0-1.4.1..704
-1.1,1518-1.5/10143-1.1473090.216918
-1.365029-1.145601
1.00000
-01
11
0.1 .l.?1430. s I hlt. 1
-0.1',5421-0.12,.071-0..2:340.:10-01-0.1072420.223500
- C.9.50730 -010.1413730.8492441.00000
12
-0.191421-0.24'.so50.2.194640.1109430.4444670-010.119112-0.1887550.4520420.164491
-0.654721-0.605660
1.00000
13
0.6231160-01-0.81,1,7) -01-0.9244490-010.5J75510-02-0.4581650-01- 0. '120n/50 -01
-0.8286830-01-0.417623-0.351J42-0.135100-0.424193-0.203535
1.30000
14
-0.119302-0.1,.77130.5568510-010.4696040-01-0.6360100-020.7661800-01-0.119168-0.121094-0.5716450-01-0.360329-0.4260270.165)110.4653011.00000
15
0.476550-010.5145/J-010.5/.301/0-01
-0.2197160-010.6106J8J-01
-0.3404660-010.3060110-010.6952270-020.4/97/70-010.704L100-010.7764551) -01
-0.2221120-01-0.7993740-01-0.5023780-01
1.00000
16
0.1168070.186690.113035-0.3148650-010.5169180-01
-0.4227010-010.2353600-01-0.1120533-010.1507150-010.7051230-010.6720363-01.-3.4081c57 -01- 0.6560473 -01-0.5768320-010.8o5720*1.00000
CD
Cb
CDto
VAstAOLE
hir5fR
17
19
20
2122
23
24
2
'.1.
+.5,6810-02
).6751166-01
0.3961470-/1
e 134 167
0.67943-01
1.139171
:4=00-01
ti,3-01
0.5/091;00-0
-0.2117150-01
-0.145066
-0.267101D-01
6.3525370-01
-0.3349190-01
3p.1 7
"19
1.45
'10.
1%-'n
0.67005)0-m
0.1(.140A
-0.121509
-0.122169
0.786010
-0.812353D-01
4-V
.3&
6-q1
"-nl
-0.1
9.4.
11'1
q-.1
1-0.7s11929-01
0.7122750-17
-0.0559/90-07
-0.0015450-01
0.4636000 -01
0.9517860-02
-0.,9941q0-01
-0.e051749-01
0.415114) -01
0.93)015
-0.42v4300-01
-0.4378160-01
-(-.4.17xo0-)1
-r!.24)074')-01
-0.47)0210-01
0.226143-01
-0.6666490-01
0.5119400-01
-0.2192230-01
7c.10-4,46
0.)1(1)94n-ol
-0.17/:131.)-02-0.660690-G2
-0.1010563-01
-0.291741D-01
-0.7142593-0/
ean;4450-l1
c.1971 1i0-01
0.217'.4.11' -11
').1739)..0-01
0.21
01:,7
0-4.
11-0.74(zeg1r.)-01
-L.1143,19
-0.104144
-3.4!/1010-01
0.1354143-01
0.4.145100-01
0.263302
0.9.'611420-01
-0.46;146
-0.340461
10
0.95..,460-01
1.2-.??0-01
0.454/3'30-01
-0.2219140-01
0.Z90430-01
0.304/G217-J1
-0.22's300
-0.160553
11
1.411314n-01
0..1113690-31
-0.1307410-01
-0.2556740-01
0.5591660-01
-0.116113/
-0.647158
12
-1.6,o6m-11
-0.9640010-01
-0.3369450-01
-0.435.3440 -01
-0.5657100-01
0.1310?..4
13
-0.110150
-0.4:;536:0-01
-0.100760)-01
0.190580
..0.2034/6J-01
-0.119667
1..141)M:
14
-0.107001
-1.5540510 -01
-0.7414450-01
-0.4576970-01
0.726645
-0.1352906-01
-0.6602910-02
0.540611
is
.455.740
1.6349100-01
0.5086140-01
-0.152605
-0.1)7660
0.801492U-01
0.232426
-0.5203140-01
U.
C.587782
0.941570-01
0.116470
0.205992
-0.229741
0.8161170-01
0.338460
-0.6826110-01
17
1:00010
q.tosR10
0.118296
0.278992
-0.156088
0.8174800-03
0.515916
-0.116014
IR
lonno
0.932854
-0.5051423-01
-0.215710
-0.14228e0-01
0.8293540-01
-0.8176600-01
19
1.00000
0.6339270-01
-0.238449
-0.2077/90-01
0.9295133-01
-0.9180560-01
20
1.00000
-0.106229
-0.6240010 -01
0.236155
-0.2206770-01
21
1.00000
0.2619270-02
-0.155997
0.204750
??
1.00000
-0.5302620-01
-0.2207780-01
23
1.00000
-0.113753
241.00000
V3v1481E
25
26
27
Nu1RER
-:1.
6544
050
-11
1.4,
,,,1u
n0.418757
2-r.16114,4
).36s49
0.1)1141
1-c.25c-941n-c1
-).
25)^
17-C.2701/9
4r.3.18.)730-11
-1.141290
-0.149193
-P. 17554c0-01
-1.3355410-01
-0.416i100-01
r.1%2:440-91
-I.111167
-0.163982
7_0.13/.121
1.221114
3.21
6699
-r.)
.., 1
:74
-1.1
/191
99-0.210977
9 11 11
-r.%.,,,q4
-r.t..1..%4
-0.41ce..4
1.41047140 -01
1...2v.10
0.44.r.17
0.1::2020-01
(7..)471)7
0.954700
12
1-.1-1-440 -11
-).6,.15S1
-0.014577
13
c.)51. 70
-1.1.1449
-0.3
VI.1
,, 1
t:
C.'.05':4-7
-1.4114.'3
-0.11967S
lc
14 17
-r.9.47711n-11
..c.,,,1:261-nt
-1-.14A-457
1.ri1 -Qc.)-01
9.7^19559-)1
1.1 17449
0.m141341) -ot
4.477570-01
0.125115
18
-r.111,,4
0.7)167,1-11
0.5964910-01
10
..".110.31,t,
0.73.78q0-11
0.640931-01
-n.)14411,-11
-9.165sJs0 -9:
-;'.'441;.40-17
'1 2?
0.194(1)
--0.1?.170n0-11
-).-1/2:11,:...0-92
9.4:11410 -01
-n. 3.1.8 ^0-112
0.4717111-01
23
-C.79,613 3n-11
-0. Iscr;S2
-0.174707
74
C.964)&9
-1.163465
-0.361020
?s
1.00000
-0.545856
-0.530493
76
1.00e01)
0.985853
27
1.03n00
0 0 7
393
All Occupation Strata Black Males
VARIABLE. MEAN ... - STANDARD DEVIATION
SCHOOL 1 8.512061404 3.414289615
WAGE 2 2.366831140 0.9085147540
UN MEM 3 0.4539473684 0.4981478202
MIG<50 4 0.4265350877 0.4948448302
S.AT16 5 0.7182017544 0.4501223774EXP 6 29.92872807 12.63184550
TRAIN 7 0.1304824561 0.3370182869
PHY DM 8 2.773026316 0.8169556398
NEG WT 9 0.6918859649 0.9650420761
GED 10 2.763157895 0.7470116868
SVP 11 3.703728070 1.697336754
%13MCLC 12 0.1239243421 0.1310786869
%WFOCC 13 0.1514462719 0.1745036535
%8FOCC 14 0.22194C7895D-01 0.40550880730-01
DA/PW 15 22.68336428 40.60823057
NYL/PW 16 2.870881390 3.370130879
AVEMPF 17 0.3652142325 0.2873401759
FDSL 18 0.41644143720-01 0.92480274620-01
FOSLEX 19 0.66097931060-01 0.1016559490
ATPFAS 20 0.40800438600-01 0.18882315060-01
MININD 21 0.3101666323 0.1631280815
EDS16 22 5.646929825 4.563049152
UNXMPF 23 0.2229174342 0.3135152414
%FMOCC 24 0.1736403509 0.1997085858
M1NOCC 25 0.3025646930 0.1977858175
GDfSV 26 6.466885965 2.394528147
GDXSV 27 11.38012061 8.195603296
411
CORaELATICh MATRIX
V4R14eLE 1
N0M8ER
I 1.00000234567
2
0.2898511.00000
3
0.6325260-010.4313941.00.00
4
-0.113622-0.226186-0.131745
1.00040
5
-0.303842-0.299593-0.192565-0.8565320-01
1.00000
6
-0.579347-0.1005840.8,110580-02
-0.1989210-010.1704081.00000
7
0.266210.195230.19486 3-01
-0.10371-0.14085- 0.25308
1.0000
8
-0.177485-0.8258170-010.5325820-020.63246;C CI0.1303510.1321.82-0.5577650-01
$0/
t"
3VARI:.5LEisa.m,Eft
9 13 11 12 13 14 151.00000
16
N 1
2-3.9664930-01-c.I;(1_4..)-:;1
0.2165980.113558
0.1921720.2.;0.:51
-0.225329-0.21805 g:2=-01
0.263',660-01-0.124694
-0.613701 -010.762397 -01
0.5631090-010.200415
3 C.2411063-01 -0.157055 -0.9664550-01 -0.4148190-01 0.2434900-01 -6.2849130-01 -0.384756 -01 0.213670
4 0.1116290 -01 -0.5365440-01 -0.5651720-01 0.7670293-01 -0.9370050-01 -0.2o23000-01 0.200012 -01 -0.1563610-01
5 0.2227473-01 -0.138967 -0.116638 0.111489 -0.4265550-01 0.2903550-01 0.592703 -01 -0.6706573-01
6 0.7959900-01 -0.151812 -0.145132 0.150112 -0.6483409-01 0.2527873-02 0.751419 -01 0.1133090-01
7 -0.3487.140-01 0.120491 0.123875 -0.136442 0.578100-01 -0.1093130-01 -0.537744 -01 0.3246000-01
8 0.592040 -:N.34233d -0.187319 0.413955 -0.,66432 -0.193799 0.675164 -01 -0.1617340-02
9 1.00000 -0.193181 0.2804000-01 0.110629 -0.244834 0.104109 0.575852 -01 -0.2803590-01
10 1.00000 0.904934 - 0.501733 0.2005430-01 -0.140.399 -0.654/01 -01 -0.5048520-01
IL 1.00000 -0.483609 -0.11%937 -0.144145 - 0.59s973 -01 -0.5429370-01
12 1.00000 -0.374331 -0.7725220-01 -0.606504 -03 -0.9937010-01
13 1.00000 0.550263 -0.715793 -01 0.1694060-01
14 1.00000 -0.100166 -0.128079
15 1.00000 0.644337
16 1.00000
vs.:At:.1 17 16 19 W...49tK
1 1.15512,0-Ji a.J.3,e)-01 0.7aw56?,..3l 2 c;.,-3;sJ) ,/ '" 2: C.:36 3 J ..5 P.-.:.)-0 1 ^.. 34 14c.11) Ui 4 -0. 1J3, 1J-01 -J.6C1.-.'1-01 -u.,.432910-01 5 -3.1,,,4:6 -n525,,,%:0 -,1 -;.6...t.,-,1'.3-Z.1
6 C.:0.1940-01 0.6.-8"10-02 -c.172!,:10 -01 L.;J1,3...-01 71.0:,.,,,)-;-:( 2.114617
8 C.4263-160-01 0.75,,i30-01 0.49910-01 9 3.461'26:.0 J1 C.56 :6'50 -01 .1.'1'.451C3-01
-0.8819940-01 0.1-170-0a n.q23770-31 11 -6.255721v-01 0.145945 0.161354 12 -C.7642500-J1 -0.105.'25 -3.1175,,2 23 -0.3667210-01 -3.610:1%0-01 -0.7765830-01 14 -:.165394 -3.822673J-01 -0.107541 15 2.173551 -0.25794,1.,-01 -0.45;1.150-01 16 0.681092 0.29525)1D-01 C.6,11733-01 17 1.00000 0.135232 0.203351 18 -.00033 0.,;76992
19 1.00000 j 20
21
CA) 23 22
24
VASIA6LE hum at 6
1
2
25
-C.10e123J-01 -0.145744
26
9.236)10 0.1 :91.35
27
0.201059 0.214:369
3 -C.1195170-01 -0.1177E3 -C.126191 4 -0.3122040-01 -0.6391E00-C1 -0.6t42650-01
5 C.422,):43-01 - .3.133119 -0.131832
6 0.3111440-01 -3.152419 -0.141551 7 -0.4176540-01 0.1.'6052 0.133292 8 -C.265148 -0.239577 -0.199117 9 -3.121175 -0.4137700-01 0.2516790-02 13 -C.347:".8 0.953419 0.v41797
11 -C.451±46 0.991149 0.92q151
12 C.316625 -9.491365 -C.46)979 13 0.747023 -0./523710-01 - 0.9531820 -01
14 0.839516 -0.152371 -0.155226
15 -c.a,csliu-01 -0.457,v0J-01
16 -0.7521620-01 -0.56/3C9D-01 -0.46:5.2D-01
17 -0.116748 -0.3951560-31 -0.2772060-01
19 -C.160115 0.1254C') 0.123992
19 -C.181752 3.143193 0.139971
20 -C.7106570-01 -c.'.5393S40-01 -0.4855070-01
21 0.150163 0.2286950 -01 C.1101010-01
22 0.3422390-01 -0.1571610-01 -C.1:96!,40-01
23 -0.6147013-01 -0.9111940-01 -0.907s473 -01
24 C.762555 -0.9666040-01 -0.114607
25 1.00000 -0.428536 -0.421426
26 1.00000 0.983491
27 1.00000
20 21 22 23 24
0.14Z:82 0.637:,:73-01 0.296829 0.53/04 I-01 0.137166 G. 35c,',46 -0.25-01v -0.112,.12 0.46325 -C.7263520-01
3.432' 13 -a:34264 -C.122577 0.76J25 0.154.01 -0.3,::354d-01 0.:.457..10-02 -0 151693 -0.8;a :le )-ca -0.872,6,0-j1
-0.171,.'59 -0.31128t0-31 0.775606 -0.15z:6 -0.3117550-01
-0. l'e1061)-01. -0.4.))4t:1,-GL -0.111::,7e 0.52825 J-01 -0.7361630-01
0.1):522 -0.1015350-01 -0.04135'10-02 0.5',264 3-01 C.451v030-01 -0...5216:3 -01 -0.150560 0.50',16-.-01 0.558,/6 3-01 -0.534294 -1-71/4160 -01 -0.10c729 -0.1151970-01 0.51680 3-01 -0.192:,19
-0.52,033c-o1 0.46.'0230-01 -0.144i56-01 -0. 2:.24 -C.1514/00-01 -0.....:5112..-01 C.1166523-01 -0.15:0270-01 -0.42984 3-01 -0.1e,(20
-0.123714 0.3937620-01 -0.1960663-01 -C.57492 3-01 -0.342773 0.437:290-01 0.231177 0.4012.:3-C1 -0.62470 0-02 C.961.522
-0.13:039 0.426120 0.:7:06(.,b-31 -0.96160 J-01 0.683865 -0.392:330-01 -0.220149 0.2114010-:11 0.10002 -0.6268410-01
0.540753 -0.120109 -0.3120160-01 0.4/v/2 -0.9456120-02 0.;4Le96 -0.321539 -0.5796820-01 0.73644 -0.6560183-01
- 0.131368 -0.233201 -0.1761540-01 0.66933 0-01 -0.6523630-C1 -0.2663653-01 -0.263955 -0.4035420-01 0.13165 -0.6969330-01
1.00000 -0.259840 -0.8242250-01 0.51801 0.1.:63473-C1 1.00000 0.1962.100-01 -0.29359 0.230544
1.00000 -0.10930 0.4076020-01 1.0000 -0.2512420-01
1.00000
396
All Occupation Strata White Females
VARIABLE
SCHOOL 1
4AGE 2
UN MEh 3
MLG <50 4
S.AT16 5
EXP 6
TRAIN 7
MEAN
10.574034332.051384120
0.26716738200.47961373390.283261802629.59314764
0.1083690987
STANDARD DEVIATION
2.3160328570.98395202940.44271804980.49)85246040.450824384110.58284685
0.3110128995
PH? DH 8 1.763948498 0.6966308416
NEG WT 9 0.1652360515 0.5228954011
GED 10 3.308369099 0.7754107196
SVP 11 4.160836910 1.366598869
'AB90CC 12 0.2644849785D-01 0.3409739609D-01
1401FCCC 13 0.5430901288 0.2622150920
ABFOCC 14 0.2999356223D-01 0.4426762)860-01
EA/PW 15 15.34410537 35.00407986
NYL/P4 16 2.2232516/4 3.017452709
AVEMPF 17 0-3503766452 0.2717903175
POST. 13 C.2486631804D-01 0.3975619450D-01
FDSLEX 19 0.:3324286530-01 0.60343967420-01
ATPFNE 20 0.40006974250-01 0.1697161378D-01
MININD 21 0.4115755591 0.1995394110
EDti16 22 2.t.,61311159 4.906322707
UNXMPF 23 0.1069819027 0.2395824421
AiMOCC 24 0.5730636910 0.2682517203
MINOCC 25 0.6065321888 0.2512549256
GD+SV 26 7.469206009 2.089204571
GDXSV 27 14.68225322 7.841667910
414
C:1i,F:A7108 its.Talz
VA3IABLE
1
34
5
0T
8
1
1.03000
2
0.1823341.30000
3
-0.1905870.915420-011.30000
4
-0.2096260-C1-0.3352450-01-0.4086750-01
1.00000
5
-0.2627880-01-0.130686-0.142787-0.363114D-01
1.00000
6
-0.416514-0.1497043.7366960 -010.1173100-01
-0.4899890-011.00000
7
C.1894120.112170
-0.7008250-0-0.6S23000-010.25913.3-0;
- 0.1621671.00000
8
-0.331916-0.2379790.201164
-0.1383063-010.1055903-020.141753-0.189121
1.00000
URIAJLEkiln:71.3
9 10 11 12 13 14 15 16
1 -0.172.21 0.416291 0.305169 -0.283748 0.163575 -0.22'4983 0.116575 0.9048750-01
2 -0.121725 0.110082 0.121708 -0.127055 0.124374 -0.14/1c8 0.174371 0.279424
.3 0.1325203-01 -C.259012 -0.25447, 0.144464 -0.6258890-01 0.643301D-02 0.3551343- 1 0.8012133 -11
4 0.2112393-31 -0.2741160-01 -0.5379073-01 0.1662390-01 0.3,,67343-01 0.5142533-01 0.3510920- 1 0.5425710-31
5 0.4726500-01 -C.5740733-01 -0.04.5243-01 0.6134753-01 -0.2774263-01 0.654211D-02 0.902V.4a- 1 -0.442(.923-)2
6 0.2.20870-01 -0.134413 -0-S36116D-01 0.20092CD-01 -0.7916630-01 0.9266333 -01 -0.6908723- 1 -0.6911520-0'
7 -0.5028100-01 0.164147 0.149107 -0.130789 0.120380 -0.7410110-01 0.9405180- 1 0.7494063-01
0 0.502175 -0.533465 -0.316718 0.584633 -0.455619 0.201594 -0.190504 -0.127754
9 1.00000 -0.232299 0.266'360-01 0.457181 -0.269653 0.447154 -0.681.293- 1 -0.22144.20-01
10 1.00000 0.652250 - 3.630927 0.194563 -0.347398 0.6583360- 1 -0.2139363-01
11 1.00000 -0.425060 0.8737770-01 -0.255182 0.6751220- 3 -0.4768113-01
12 1.00000 -0.602695 0.259836 -0.7557313- 1 0.1140720-02
131.00000 0.5342530-01 0.10b733 0.2200210-01
141.00000 0.69567i3- 1 -0.119156
151.00000 0.763602
161.00000
vsalx0Lz
17
18
19
29
21
22
23
24
1D. 1z6.21
C...1',5t,,90-01
-0.2c.1-1::.-02
3.4'1,993-01
-0.1398)7
0.121,10
-0.300633-0;
0.12)411
23..n0.:73
0. 1.; ::+33
0.12It,/
9, I.ilv II
-0. 235',20
- 0. /71679J-01
0.170212
0.7729233-01
3. MI 145
0.6 :7 /4C.,- C1
0.13'7745
0. 17..;2 71
:1.7).
)1.2-02
-0. 1 )13 3,
0.119941
-0.6318495-01
40.13445)0-02
-0.4..,9994.;-01
-0.5219130-01
0.221151 -01
0.1,16:):..-01
-0.q/1315D-31
-3.717-33 -02
0.4'),)9930 -01
3.11-4713:.-01
-0.546_1:1)-C-1
-0.9.:16170-01
-3.e.',075D-01
0.3:,/:217) -01
0.902Iil
-0.2.10.67D-0'
-0.25)422D-01
6- 3.134221
-.0.3.9otC::-C1
-0.50:',C..-01
-0.c.:34%53-01
0.131139
-0.1!:215
-0.1;e314.,-01
-0.5:10(230-01
7G.e38.1,11 -01
3.1,7 t4 ::, -!:1
0.22730-02
-0.45.1140-01
-0.115111,
0.33:G040-01
0.'.o76343-01
0.105436
-3.12..364
0.':.:.4::.-:.i1
3.212742
0.t17...:43-01
0.1:,):29
-0.494'3d03-01
:).7%c974a-0'
-0..11753
.4
-J.'19',4D
0.51110,.$3-02
v.1712.60-01
-3.4)23403 -01
0.7324583-01
0.11G3J33-2,2
-C.1217172,0'
-0.1,J1..09
10
0.66221.93-01
-0.1.1475
-0.217312
-0.11G991
-0.4057180-01
.0.001N:11.7.-02-0.)6d0,1
C.132dA
11
-0.1305483-01
-3.419452:J-A
-0.127123
-0.105220
-0.3453200-01
-0.4253231-01
-0.170647
'
0.3324860-01
12
-3.57J-133D-31
0.5,82363 -01
0.190639
0.118203
-0.2252100-01
0.59.!31D -02
0.62070-01
.0.544234
13
-0.4131663-01
-0.3619560 -01
-0.147931
-0.232816
0.3422220-.01
0.3323510-02
-0.6500633-0'
0.506.117
14
-0.204047
0.34o09 :0-02
--3.5334747-01
-0.179303
0.172097
-0.346204D-01
-0.651:3923 -01
0.217320
15
0.517723
C.6669150-02-0.232o9D-01
0.7415663-01
-0.242261
0.6452392,01
0.226243
0.9285040-01
16
0.614972
0.714+9310 -G1
0.146962
0.436251
-0.311494
-0.1243V:0-02
0.312075
C.1634543-02
17
1.40000
0.2005673-01
0.I31111
0.359369
-0.244637
0.4723410-01
0.460363
-0.6256173-%,1
Id
1.00000
0.045453
0.122752
-0.206834
-0.460380-0:
0.613%260-0:
-0.33979,0-01
19
1.00060
0.354038
-0.335435
-0.8663030-01
0.123064
-0.152963
20
1.00000
-0.443967
-0.4094900-01
0.262425
-3.2.7854
21
1.00000
0. 158614D -01
-0.178:131
0.51864.9D-01
22
1.00000
-0.1269700-0
-0.246833D-02
23
1.00000
-0.9390553-01
24
1.03000
VIRIASLE
bj:0.1Ld
1 2 3 S
25
0.932517.1
0.t1057150-01
-0.4451750-01
0.5/3375J-01
-0.192.3733-01
26
0.301C;11
0.132770
-0.2G0142
-0.6462930-01
-0.709ooD-01
-0.1139o4
27
0.3E4036
0.142649
-0.276877
-0.6700640-01
-0.7t4359r.-01
-0.131..16
0.9431)2D-01
0.113422
0.154605
-0.3t0255
-0.433573
-0.3.399G3
9-0.14.313
-0.6845553-01
-0.5211150-01
10
0.5511303-01
0.93763/
0.12)2:38
-3.1-!4)630-91
(.9-13970
0.)72c11
12
-0.447415
-0.5.9920
-0.53971)
13
0.971247
C.11)2rE
0.b919:20-01
14
0..b72e4
- 0.2940 7d
-0.403970
15
0.081120 -01
0.2-37130-01
0.2242313 -01
16
0.2130 4 13 -02-0.4356180-01
-0.2991.3D-01
17
-0.9031)6D-31
0.2:1021D-02
0.2335603-02
lo
-0.2d97423-01
-0.6434520-01
-0.61213')7 -01
19
-0.14263
-0.165145
-0.153464
20
-0.227415
-0.111256
-0.9169913-01
21
0.429)350-01
-0.3652850-01
-0.9135200-01
22
-0.102632D-02
-3.2631133 -01
-0.2383463 -01
23
-0.4995473-01
- 0.175782
- 0.171670
24
0.593518
0.7o04563-01
0.4071177 -01
25
1.03000
0.8695410-02
-0.2571530 -01
26
1.00000
0.990209
27
1.00030
O C 0.1 rt 0 Cr)
-3
CJ rP 0 0 7
ARLABLE
399
All Occupation Strata Black Females
MEAN STANDARD DEVEATION
SCIICOLWAGEUN i':E311G<50S. AT16EXPTRAIN
Y
NEG hrGEDSVP
1
2.3456789
1011
9.6700251891,602972292
0.3221;1813600.142569269520.672544080627. )319b992
0.13:102015112.156 17 1285
0. 70277 078092.7262115873.470277078
2.9248902680. 6u582605720.46799 183760. 49`,07 150410.460)1772956
11. 76(.,565410. 4 321. -52930.6101,'.j22450..3:1111.."176645O. 7 1 3', 12068
i. 3.!4(.0 139812 0. S2 36164967D-01 0.417 )4J.33 76D-0 1
;;WIGCC 13 0.4609204710 0.2110161055%13POCC 14 0.127022670 0 0. 1'0,001540CA/Ece: 15 9.718452352 d.i670229NYL/PN 16 1.2617 56694 2.,,./07406A V 1:11 17 0. i30.:44:0655 0. 24/02::3:i82FDSL 181 0,2576..4 ).706D--0 1 0. -..h,1:1325D-0 11: DS LEX 19 0. 4625'170t; 120-01 19J 020-0 1
ATP 1.5 2.0 0.3273 1'3149D-01 C. 1t',0 94:lb D-01:111;1IND 21 0. 544404u008 0. 1/2 /c00116E0516 22 6.272C4J3:)2 5..).13.156382
N ;11) 23 0.0613054660D-01 0.2 1572 12631OCN: 24 0.5879521411 0. i')/.;408:388
el N 0 CC 25 0.6438367909 0.24.!:i419056t:D+SV 26 6.1061468665 1. 111.0.13046
GDXSV 27 10.27624685 6.5 )7;128941
417
4,LU)
C0000t
t8ZLZS'0
L9/2,17'9-
000001
tO-COLta.b0-
C0009
9tSt
hl
Z0-CE6CSEtO-
te-CC90tte-
tO-c7C9FLt0
.S7tt43-
10-C99St760-
t0-ClIZELFC.9-
cone't
tORSttl.0-
00000'1
Etzt
t0-(76t9t.07'0
t0-GLE9£070
Si::96C0-
70-09Prolt'0-
107t.6t0-
00000't
tt
401211'0
VrEtit'0
SfoCPt0-
C7S.:,9t'0
siczt:*e-
SOSSsVe
000CCt
01
SC7197.0-
09660-
,7r90i,0
6ACPPt.0-
607-LSt0
OIL6Ct0
176L610-
00000%
69.7.:Ct'0-
OcCett*O-
10-CfrSSL^0
'z',0:(q*O-
10-:IS*0
10-CC1e596'0
ttbt-CZ'0-
e6L(LtC
PL6,91.*0
Ch99StC
t71-r9t9-
70S07tc
7.3T,it'il-
10-CVq:.5 r,()
67160
Lf9StIt0-
L909sc0-
tOcet:c,OftC-
G6E7c7C
loUtt*c-
ZSCLLt*o
t0-CSIStf9v-
Q9c0St'0-
7tec6tc
9
C9P-C7-0-
659551.9-
ISS4qt0
10-UtteciO-te-GE106S0
L0 -crs,77t.8.0-
0t7L71)-
Z7,IL0
S
to-acterso
,
LG;csetn4vo
t0-cPvmz*c-
te-cvlogzs*o
to-apqrztzc-
t0-arbcz47c
10-(16,92790
L0-Ctti.toS*0-
LO-CL/ n(P;'0
10-caeLs2*0
10-c*:"5L.0-
t0-cpcv;,:c0-
t9-cOtet7.0-
9111,71 0-
S.:95,,1*0-
01)7p10-
SttStt.0
[0:,7tr0
66.14'0-
to-C9P9$:c5'fi-
5.476Lt.0-
;57101'0
/c1^:CUO
16:SIT'ts-
7
0617C
10-Q6SLbS.0
11;.9qs-0-
10-C1.7tCt'S'0
,99tPt.0-
791SW0
C'g:£97'0
LCS-Plt'J-
1
0p.
91
09000'1
St
tt
OSE7W0-
000001
999C0
L9P09U0-
0000rt
t)
PSL*7:'0
10-0*6,(91-0-
to-cr057re*0-
t0-C7SS7.6S*0-
to-C60[S60*0
tc-c6rv.ot*0
to-Gc9(.669'0-
tO-Ctah9S.0
10-06Y;991.0
tt.t.7c7*0-
ZECtet0-
LSOtr'e,
t9SP^Z'0
CT,SSFV0-
7t(t9S*0-
99
ft
00000 1
Lr6SviO-
SSLC41.0-
LLL.97.*0-
tS9but*O-
S
tt
00000%
Lt0/'0.*0-
00000%
cotp71.0-
ttcoor.o
V671t0- 10-SEL9Z0S0
OL
30e9ct
Pit:01'10
97/..31:7*;
L0S7
OC
OO
I
".::31:r4
11,:itttrilt:v;;
C 0 U
:3 U 0
00000'! ESt Ot't '0- 00000'1.
Z0-CCt t ZWO tO-araCt Pc16'0- 1'PLErr *0 bl9tRi *0-
i! L.17;"0- Z70ZsE "0
000707."0- 005991 '0
Z0-0r6(_,5 ", *0- I0-C(.?SgOn '0
Ct 9t7. "0- 01771 S*0
l.W,t '0- 90LEEE '0
L6L .0 96!, t Ft '0 ez 91 / ot,PtIP1 *0_
FS(.79C0 t0-0£ trIT;7 *0-
/ISS9th*C- LEI-Clot S49 '0-
Ll/q cE*0- 1.0-C.+10',011'0-
7LUPt '0- 70-c.;,t(079'0-
Z6Sf 't '0 tOt 9E',' *0-
CSF CC! '0- L66771 '0-
ZO-CEPP1 clf'LS1. '0
L0-rlEf 7:61. '0 t 0-Uf 17 t Lq '0-
E0-Gt P6S'E '0 170001 '0-
tO-C!.9! 0,Z '0 t 0-et ELrY, '0-
oZ t '0- qcoq,-9'0 Sa1707.0- SoCe;1 *0
t .0- i;1 *0
172 rZ
00000.1 Ilt1:f6G0 00000-1
Et 6E67'0- it( 0G300't 01711-2: 0- PP.O*7'0- LOL "(
t0-0Li1101q.0- t0-,11%r S0t.6St C- 10-0/.1*):1E-0- tC-.CPP7qCf'0- EO-CZ-Pf7V1.
ZO-CtLo,7r 0- *0- 767:SF 'f 10-C3n1::C0 10-0'L S16'0 S,.-
70-0":"EZEt *- t0-rf.1-0,,!,1 Ot 77 *0 t 0-0q1'4,t h0- 10-01,/0,:ti 0- LC-V.-",Sit 0- t 0-(17'tt`..t.'0 to-PA-1f Jr10 07t *C.-
lCI-Gt!i(Lf 0 te,-C/ r P919-0 rf.tiPt '0- to-Ott t:0190 10 -1lotrUS *0 L 0-Cif 0(.01 0-
0'19t.PE "0- I07rt,f10- to- c.)0St CZ*0 10 '0 *0
-.1r712.0_ SLS)17 0- 9-E,Q- '0- <417r,CP
f C trc-LF,NC Flt 7 Z1- t '7'0 -rt cz.vf- Sot LF7
7C-r71:.:660" /0-1-911 icr"G- t t - L0-pc,p( ci)L *0 t0 -fr.q..,tq -0 tO-C.LIrttr "0-
140-a,;et,"tc0 '0- f,'"?( ;C: *0- to-(:ct "t t "r`
LC -"M "1:-.160- 0;1 0- Sc,r,COt' C
Cf f-t" nC+ ti-11:7t,7:;00 10 ror P ti'iSt *0- I t n- zrzfit 0- r CC,1%71 0 1rott71 -0 t^ 172"f
76(10L1'0 t5r,',6t'0 hn cf7'0- t
L.? oz
FZ
t C7. f
cl
9: St of Fl 71 tL 1
L
L7 97 57
00E00'1. Z0-06Zt7Lh'0 CO0C0 "I.
to-CEIS(stV0- f 1001S.0- 00000'1.
Z0-0SL L9C6 '0- ZPLSfE*0- SL9tOr*C 00400'1
to-CP67Zth '0 1Z IS:7"0- 10-01,00LbZ.0 Of t SS0 "0 ' OC.0' I
SLOCt '0- 77cPt17'0- 0'7E105'0 OtbLSZ ' 0 I0 -C' T.F h' 0 ECDCC't
'r6S'St '0- $,St'i 61'0- FL It CC C OEW-5t "0 10 -Ck 4',V.,10 6cw'Orq '0
6t7SE1 ^0- 0900E7'0- t0-01ELLP1 '0 10-C1i ^0tS" 0 t0-01976.11-0 e.,.,,t:t,'0
IC-C[SfUZ '0 6t E717. '0 67CF,'S '0- 7",£17'0- L0-r0ects-c- 0c,.rS7";--
t0-e-q 0 n -0- L'f.777 *0 7S(ELt '0- t0-Gloo99L '0- l0-,:f i'f ;71 -0 t 0-Cc.770LS '0-
t 0,-(1( Ct St '0- 10-06'6U- l '0- :,,tott 'E IJ-076P5( ""0- t0-'7.,O. 0111'4' r- #-17'.t, t *C.-
i 0-1',S t t7E '0- Z 0-00 ; S6P '0 10-C° t,fl' :n-tict )1C t '0 t^ -r-,, ''tt.'.1- LC-0( c:.nt 'C
v)-r(A;mzo- to-ozpvq h'e lcZr ^t '0 10-a6q1u0 -0 t0 ....,...;:- I -0- t 1 i 7,;:
10-0£!=('7( '0 - E7'f 7'0 Sq°,,A 'E- IS It .+70- :" ( ;7: .0- .-,,t ,f .r-
to-.-r , r.,-,*o te-otozcece- to-lt",$)N 'F- te--(..,,--,zi.-. I.--,/.1:!t. -0- Lit "c' I z-
to-nGetoco Vt. 611 e- z I t.'-tt *C. 6atft '0 t^-eif -1"/.'0 ( i7nt 7'0
f PCcSt *0- t0-C97:"..1, 0( q.t. 4..Y.St 1 'e- 10-::t1,1 %,-: '0.- (4/ '9: '0-
nctt :P*o t0-(39!,/tcce 7gE 1 1 Z "4-- t0-6nStc60- to -:ctILI t *0- t( ct;,t *0-
OtC1ZE '0- 70-E7(0059*C- t0-Cq9elbi '0 l 0-06t 741., '0- 1 0-lt ",e '0- IC-E- I t f 0E 'C
66f; Cl 0- Lt 6,f t '0- lq St t- 'C E;;7601'0 1.)-716:9t9'0 L6if*c:*C.
10-01f-75'1'0- SiflIt'o- if t..,t.= '0 1.t t'r.c,!' a 7tqfts.") I tt 6.,,f 'E
St ^q97 '0 rt Z'c',Fl *0- F j...-.7'E crZ5(,t '0 13- Vt t 6C'y 'C is4 ".:7*C
7.7 LZ CZ t,t
livVIPYA
EF
tZ Oi
6: Pt Lt 4, St wt Ft 7t tt rt
t
S
402
All Occupation Strata Cross Race-Sex
VARIABLE mEAN STANDARD DEVIATION
SCUOOL 1 10.77109440 2.9100284642 2.953421053 1.517826903
UNioN 3 0.3642439432 0.4813180476MIG<50 4 0.4298245614 0.4951543409S 16 5 0.2861319967 0.4520462749Exp 6 28.37176274 11.62024710RACE 7 0.8604845447D-01 0.2804941750SEX 8 0.2878028404 . 0.4528333252TRAIN 9 0.2017543860 0.4013936397Ruv Dm 10 2.210944027 0.8625740317NEG 4T 11 0.4482038429 0.8344037001svp 12 4.853968254 1.878014896.5.03mocc 13 0.4767126149D-01 0.5756481215D-01.4wFocc 14 0.2745614035 0.2727570487
A3FoCc 15 0.190296574 8D-01 0.412437683813-01
DA/pw 1t 27.63948124 65.831007.86
AV P? 17 0.3995127541 0.2919289129Fasc 18 0.3538361421D-01 0.6429166881D-01AeppAs 19 0.4308847110D-01 0.175878961713-01
XmINIu 20 0.3423889435 0.1894500734EDs16 21 2.837510443 4.806049936
uNxNRF 22 0.1734650654 0.2933291042:4 i1GCC 4 .) 0.2935910610 0.2873789668NmINoc 24 0.3412623225 0.2748465174aAxsEA 25 0.25480367590-01 0.1576118492
420
C0RRE..LTIUM MATHix
VARLAJ!.?WJKSi.4
1
23
67
8
1
1.30000
2
0.3608b61.00000
3
-0.1028530.1702123-011.000J0
-0.8714660-01-0.1536660.595450-011.00000
5
-0.165901-0.15.053-0.120053-0.6614590-01
1.00000
-0.457394-0.1304610.8127710-010.4070460-010.297107D-011.00000
7
-0.150629-0.1605760 2156100-01
-0.1636363-020.2210010.409060D-031.00000
-0.4924353-01-0.40.412-0.140280.7240680-010.3645000-010.6963560-010.5344490-021.30003
VARLA3LEaum883
9 10 11 12 13 15 16
1 3.234175 -2.321)31 -0.1d4t,39 ).344472 -0.171431 0.4671,10 -01 -0.176372 0.71,4920-01
2 0.1v:505 -J.7120,..50-01 0.12..00-01 0.37,1.39 -0.131714 -0.251422 -0.233203 1.155126
3 -3.506600-01 0.2?6920 0.516:0;D-01 -0.156517 0.190692 -0.155247 -0.2167910-01 0.6C,..52;10-01
4 -0.116v:3 3.3124330 -01 -0.9!;030-02 -0.7964580 -01 0.5644663-01 0.4335540-01 0.2802330-01 -0.921';70-02
5 -0.6725320-01 0.6464210-01 0.753:500-01 -3.105523 0.114181 -0.5133170-02 0.6051.610-01 0.2151r.30-01
6 -0.301977 0.115579 0.8572140-01 -0.9571810-01 0.130163 -0.2197960-01 0.7516120-01 -0..1310010-01
7
a-0.530463D-01-0.144860
0.11207-0.330946
0.9225620-01-0.186701
-3.214124-0.257678
0.301917-0.169314
-0.3816160-010.667509
0.1732020.280526
-0.4577(.20-01-0.12)483
9 1.00000 -0.3469180-01 -0.1602210-01 0.232347 -0.132/66 -0.4215260-01 -0.102366 0.6473550-01
10 1.00000 0.526995 -0.9146270-01 0.534481 -0.552900 -0.5319890-01 0.1847690-01
11 1.00000 0.113308 0.281576 -0.353346 0.0716400-01 0.5947720-01
12 1.00000 _0.517724 -0.330603 -0.331953 0.7396260-01
13 1.00000 -0.342580 0.8362900-01 -0.3440790-02
14 1.00000 0.288422 -0.9038320-01
15 1.00000 -0.7755530-01
161.00003
c.
4.J0U
00000't
SZ
OL99SL*0
9Z
9LLtItt"0
CZ
LO-CSMOS.0-
ZZ
10-0SELL9S'0
1Z
LL9051'0
OZ
tC-010199L'0-
6t
to-rrtPtcco-
el
Le-0094,*L'0-
Ll
1C-OZ.St<60-
91
91111C-0
St
LnfOt"0
hi
LC -G' '?f'0
Et
tl,C:L'O-
Et
L0-OnQuo,-0
Li
7C-C,,tctlyz-0..
01
tC-Ctf,9c-t7'0-
6
Lq(**',Z*0
P
6fSZS*0
70-060S%P*0
9
LC-CZ0St6C0
S
Z0-0,,A0P190
9
to-erNszr-o-
E
69SCL'O-
Z
te-C9r.tEPS*0-
I
67917rm
SZ
319YIRTA
000001
L10096'0
00000'1
FIZ
CZ
t_
ZSCI.SL'O-
C194Lt'0-
t0-00t597.C.0
10-0990Z6U0
00000'1
tO-OLS6659.0-
000001
ZZ1Z
NI!
05,499E'0
91Z96£'0
06SEZZ'0- L0-0ZS9LS'0
00000'1
OE
toSC7.991'0-
606SL1'0-
1PL96Z'0
10-094998L'0 -
99P7,07,'0-
00000't
6t
C9LELW0-
LOtSOL'O-
10-006P169'0
t0-ei.p0611.0-
sczazvo- E0-009hL16.0-
00000'1
P1
01'3E591'0-
SfiLLWO-
S6999G.0
t0-COtZt9t'0-
67/.6E7,'0-
7,C6LEU0
10-0SCEL66'0
00000'1
Lt
SSOZOL'0-10-0071696'0-
PCL'ICZ'O
l0-OCOCSZh'S
6C59ZE.0- 0-09C9LL0.0-
10-0660CC5*0
77tES9'0
9t
p?sr,It'o
17v7LI40
10-09FEL'0- 10-VEL9C7'0
Ph9L6E'0
,t'qECL1'0-
t0-09LSIS'0-
(69%Lt'0-
St
cq
0,76C96'0
(15061'0
Ott991.0- 10-0ZIOELZ*0
OtZt:.C.0
1Et6S1.0-
6117,01'0-
P6PZZI*0-
6t
ZS6Ltt'0-
PtIELCO-
LOLOZL*0
LO-CtLPOOL'O
6f4ZIt'0- to-t7SE;:;01'0-
t?-ccf..etl'o- i0-cofbriP'o-
it
9LZ969.0-
91069(.0-
L0 -GO'Nt)IP'0- to-cocetwo-
bst,:ft'o- to-ctLtobto
10-0.66,)E6C0
70EcC1*0
Zt
0,
POL817'0-
CLSZZL'O-
oftrnh'0-
10t,'ES':-
10-C66C:16'0
10-0,At61 :'0
ri09c7,*0
10-CLhttSi'0-
L9CEIL'O- LO-OLChLOCO-
958P91'0- ZO-OOLOWO
1.1-016L0CU.,
ZO-0/.P9tet'e
10-r6E6SW2
70-0t751517'0-
it01
Z69,41*0-
folZet'0-
70-0I;;X61'0
L0-0t9LItE'0-
10-0tZcPt6'0- 10-05N'Z'O
f3-1/716t('0
S:1/.0.0
Wth9'0
99Ctit'0- 16-r)uSq9h*0
S49694'0
cw.GS!'0-
fl,rft.0-
i=:00-
:
to-ctP.,:is'o
to-ctzgrwu-
90-0kuin'0-
it.119L'0
10-0cM-E'0
LC-0SNIW--*C-
ro-;:,qh1c-vo
LO-OEILf:'P'C-
L
L0-Ort:at'0
10-019U,OL'0-
t0-GWC9E*0
10-oCZ901/'0-to-ctuotvc
10-Q146f:39'0-
t-C,(1Eft'0- tO-OStol"0-
9
t0-CN011i,Z*0
ZO-TASLOC'0
:0-(1167119'1,-
1S/ZE6'O
L0-OU070C9*0
6791.01'0-
tit-Ch40967'0- tC-0(trcf50-
5
t0-001,416,'P
10-0:09149'0
10-39170:U0
10-067nSC0-
tO-C1rOtS9'0
ZO-0CILLC6'11-
10-09SFqW0- 10-0PLSL11'0-
9
GCCItt"0-
09tOSL'0-
Et+91',1'0
!",,,,7!*0-
S1'751'0-
vztvri..(
:0-c-v.s1;1=yb0
sopwo
t
ci:ISs75-0-
7/c-Lre-
10-0/PL9C0
L0-C%G0C.C:-.0-
9 q:16:7'0-
tr,o;c:bo
aqvico
Pff.370
z
to-cwIfitso- 10-ce1061'0
tpnct'o- to-lizczwo
10-197,):90- 1C-OKCZW0
til-Ilft.C.LIE'0
CCOSSL*0
t
F741:;'w
hZ
CZ
ZZ
It
OZ
61
81
LL
7101IEVA
BMIOGRAPHY
Arrow, Kenneth. Some Models of Rac'jal Discrimination in the Labor
Markel. Santa Monica: R= Corporation, Publication RM-6253-l.C,
1971.
Ashenfelter, Orley and Mooney, J.P. "Some Evidence on the Private
Returns to Graduate i2dueation." Southern Economic Journal 35
(January 1969): 247-256.
Averitt, Robert. The Dual. Economy: The Dynamics of American Industry.
Strw-ture. York: Norton and Company, 1968.
Becker, Gary, The Economics of Discrimination. Chicago: University
of Chicago Prass, 1957.
. Human Caoital. New York: National Bureau of Economic
Research, 1.964.
and Chiswick, B.R. "Education and the Distribution of
Earnings." American Economic Review 56 (ay 1966): 358-369.
. "Hmuan Capital arid he Personal Distribution of Income: An-Analytical Approch." W:S...1v1,-:,vtini,ky Lecture No. 1. Ann Arbor:
University of Michigan, Institute of Public Administration and
Department of Economics, 1967.
3ondix, ileinhard and Lipset, Seymour M., eds., Class, Status and Power.
G.Incoe: Free Press, 1953.
1erg, Tv.a. F.ducation and Jobs: The Great Trainin:; Robber,.. New York:
Praeger, :.969.
Bjerke, Kjeld. "Incomo and Ti,agc Distrihntions--Part 1: A Survey of
tho Literature." Review of Incol-re and Wealth 16 (September 1970):
235-252.
Mang, M. "The Privata and Social Returns to Investment in Education:
Somi- R(.!sults for Great Pil.itain." Journal of Human Resources 2
(Spring 1967): 330-346.
Blinder, Alan S. "Was...e Discrimination: Reduced Form and Structural
Estimates." Journal of Human Resources 8 (Fall 1973): 436-455.
405
423
406
Bluestone, Barry. "Low-Wage Industries and the Working Poor." Poverty
and Hunan Resources Abstracts 3 (March 1968) : 1-14.
."Economic Crises and the Law of Uneven Development." Politics
and Society 2 (fall 1972): pp. 65-82.
; Murphy, William; and Stevenson, Mary. Low Wiees and the
Workin,, Poor. Ann Arbor: Institute of Labor and industrial
Relations, University of Michigan-ayne State University, 1973.
Bowen, William. Wage Behavior in the Postwar Period. Pr inceton:
Princeton University, Industrial Relations Section, 1960.
Bowles, Samuel. "Ttriards Equality of Educational Opportunity?"Harvard Educational Peview 38 (Winter 1968): 89-99.
"Tounrds an ducational Production Function." In W. Lee
Hansen, ed. , Fduca,:ion Inr_or-e and Human Capital . New York:
National Bureau of Economic Research, 1970.
."Migration as Invc'stment: Empirical Tests of the Human
Investment Approach to Geographical MobiJity." Review of Economics
aad ics 52 (Novemier 1970): 356-362.
."UnecInal Education and the Reproduction of the Social Division
of Labor." Review of Radical Political Economics 3 (Fall-Winter
1971): 1-30.
. "Schooling and Inequalay fron Generation to Generation."Journal of Political. Economy 80 (May 1972): 5219-5251.
. "Understanding Unequal Economic Opportunity," Afir2ricati
Economic Review 63 (May 1973): 346-356.
Bronfenbrenner, Martin. Income Distribution Theory. Chicago: Aldine-
Atherton, 1971.
Carroll, A.B. and 7t stern, L.A. "Casts and Returns for Two Years ofPostsecond.lry rechn-cal Schooling." Journal of Political Eccnorly
75 (Decemoer 1967): 862 - -873.
Clark, John Batts. The Distribution of Wealth. New York: MacMillan,
1900.
Coteman, James S., et al. Eauality of Educational Op2ortunitv.
Washington: U.S. Government Printing Office, 1966.
424
407
Council of Economic Advisors. "The Budget elessage of the PresiJent,"
The____Budet _of the United States Government, Fiscal Year 1970.
t&whington, D.C.: U.S. Government Printing Otfice, 1969.
Cummings, Laurie D. "The Employed Poor: Their Characteristf.s and
Occupations," Mont111yLII)or Rev' cow 68 (July 1965): 828-38.
Delchantz, George and Evans, Robert, Jr. "Low-Wage Employment: An
Inventory and Asses:n Northwestern University, ,.d.
(mimeographed).
Denison, Edward F. The Sourcer_ofFcononic Growth in rhr United States
and Alter:tat:c.v.-2s PeEore Us. 1.,ew York: Committee 1:er Economic
Development., 1962.
Dooringer, Peter B. "Manower Programs for Ghetto Labor Markets."
2roceedig_ofth 21:;c_Aanual Win:er_Meeting of theIndustrial
Relli.Lea$ Rcsearh A,;seciecion 21(Decumber 1968): 257-i67.
.and Pi)re, Michael J. Internal Labor Markets and sliviower_
Analws:is. Lexington: D.C. heath. 1971.
Dunlop, John., 4:412.i Doterminition fluder Trade Unions. Nev. York: John
Wiley 1944.
. "Productivity and the Wage Structure." Jr leceme Fin)lovmer,t
ziadlublic Polio?, pp. 341. -362.. Edited by Lloyd A. Metzler. Ncw
York: Norton, 1948.
Durkheim, rmlle. On the Division of Lahor in Societi. New York:
Mecmit1,-,n, 1933,
Edgeworth, F,Y. 111.qual Pa.: to Men an Women." Economic Journal. 32
(Deccmbr 1922): 43J -457.
Farrnr, Doraij E. and Glauber, Rohert R. "Mult.ic_ollinearity in
RegrJas:on hnalysis: The Problem Revisited." Roviewof Economics
dnd St:tzi,itic:; 49 (February !967): 92-107.
Fine, Sidnc: A. "...no IRe of 'The Dictionary of (,)cupntionnl Titles'
As A Source of ;-;tima'..:s of Educationsl and Training Requirements."
- ioamon '..'coerces 3 (inter 1968) : 363375.
Priedran, Milton, Price Th_orv: A Provisional Text. Chicago: Aldine,
1967.
Galloway, Lowell. "rhe Negro and Poverty." journal u_f ;iusiness 40
(fanuary t967): 27-35._____
408
Garbarino, Joseph. "A Theory of Inter -- Industry Wage Structure
Var iat ion." (quarterly Journal of Economics 64 (May 1950) : 282-305.
Goldner, William. "Labor Market Factors and Skill Differentials in
Wage Rates." IndustriajRciations_Re-=earch Association, Proceedinvs
of the Tenth Annual Meeting 10 (December 193S).
Cordon, David. Economic Theories of Poverty and Unde-romplovment.
Lexington: D.C. Heath, 1972.
; Edwards, Richard C. ; and Reich, Michael. "Labor Market
Segmentation in American Capitalism." Cambridge: Harvard
University, March 1973 (mimeographed).
Green, Alfred L. Manpower and the Public Eme_ ent Service in Europe.
New York: New York State Department of Lab, 1966.
Greenslade, Rush V. "The Economic Effects of Collective Bargaining
in Bi...:JMillOUS Coal Mining." Ph.D. dissertation, University of
Chi,-:ago, 3952.
Ilanoch, Giora. "Personal Earnings And Investment in Schooling."
Ph.D. dissertation, University of Chicago, 1965.
."An Economic Analysis of Earnings and Schooling." Journal.
ofiinmIln Resources 2 (Summer 1967): 310-329._______
II nsen, W. Lea. "Total and Private Ratess of Return on favectment in
School ing." Journal of Political Econ.ny 71 (April 1961): 128-140.
_ :Weisbrod, Burton A. ; and Scanlon, William J. "Doterminants_ _
of 1-Innings: Does Sch.)oling r:eally Count?" M'dison: Discussion
Paper, Institute for Research on Poverty, University of Wisconsin,
August 196t (Preliminary Draft).
and Weisbrod, Burton. "The Distlibution of Costs and Direct
32a?Iits J: Pub3ic EdIcntion: The Case of California."
j:ounal _or_ t;umdr_ Rer,ources 4 (spring 1969): 176-191.__
Weisbrod, Bt:rton; 4,nd Scanlon, W.J. "Schooling and Earnings
of Achiev,rs." Amer icon Fconomic Review 60 (June 1970) : 409-418.
Harr:Fen', 110.,nett. Educdt on TrainintL2nd the Urban Ghetto.
Ddltimore: John Hop'cins Press, 1972.
Hay,,, William L. Statisticsfor_Psvcholeilists. New York: Holt,
Rinohart, and Winston, 1963.
Hicks, J.R. The Theory of WApes. London: MacMillan, 1932.
426
409
Charles C. and David, Martin H. "The Concept of Job
Vacancies in Dynamic Theory of the Labor Market." Madison:
Social Systems Res..aieh Institute, University of Wisconsin, 1965.
Routhakker, H.J. "Education and Income." Review of Economics, and
Statiscics 41 (February 1959): 24-28.
Hubbell, Robert and MacDonald, Martha. "A Study of Employment
Recruitm,?nt in Boston." Social Welfare Regional Research Institute,
Boston. Collnge, forthcoming, 1974.
Jencks, Christopher. ineoualit,- A R;,aspssment of the Effects of
Family and Schc,o;i-,g in Amccicl New York: Harper and Row, 1972.
Johnson, George E. and Stafford, Frank P. "Social Returns to
Quantity and Quality of Schooling. " Ann Ac her: Department of
Economics, University of MIchigan, 1970 (mimeographed).
and Youmans, Kenwood C. "Union Relative Wage Effects by
Age and Edica:iin," industrial and Labor Relations Review 24
(January 19/i): 171-179.
Kane. Edwird J. F.conom;c. Statistics and Econometrics. New York:
Harter and 1:uw, L968.
Katzner, Donald A. "Theory and Cost of Racial. Discrim4nation."
Ph. 1). dissertation, University of Pennsylvania, 1966.
Kerr, Clark. "fhe E-ilkanization of Labor Markets." in Labor Mohilitv
21nd......:ic,112portentty, pp. 92-110. Edited by E. Wight Bakke.
11.e New le-1-,nolo3r Pros: and John Wiley, .195/4.
Klein, L. R. An II:trod:lotion :o Ecene,petrics. Englewood Cliffs:
Preliticc-:;..1.1, 1962.
Kahn, 1l'/as . Tho Strue:or, re. Seiontific Vevolutions. Chicago:.....
Jni.icrsi:y .'f Chlvaio Press, .1962.
Leontief, "ftc. Strue..ture of the U.S Economy." Scientific
American Apr.I 1935, pp. 3-13.
Lester, Ri,.hard. "Short:om:Tvi of Marginai Analysis tor Wage-Employmcnt
Problems." .%:.:(rican 1%conomicf Roview 36 (March 1946): 63-82.
. "Margini,ltsm, Minimum Wages, and Labor Markets." American
Lconsritc Izeview 37 (March 1947):' 135-147.
. Mancw:(fr Plaanin" in n Free Societi. Princeton: Princeton
Univetsttv Presq, 1966.
4%7
410
Levinson, Harold M. Unionism,ya?,,e_Trouds., and JIICOLW Distribution
1914-1947. Ann Arbor: University of Michigan Press, 1951.
. Determining rorces in Co11c!ctive. Wage !in in. New York:
John Wiley, 1966.
"Unionism, Concentration, and Wage Changes: Toward a Unified
Theory." Indostrizd and Labor Relations Review 70 (ionuary 1967)
198-205.
Levitan, Sar A. "Manpower Programs Under Republican Managoment."Poverty and human Resources 5 (March-April 1970) : 4-13.
and Mangum, Car di L. Federal Training ind Won- Prmrams in
the Sixties. Ann ixbor: The University of Michigan, ln3titute of
Labor and lndustria. Relations, 1969.
Lewis, H. Gregg. Cniynt and Relative Wo,'es in the Unitnd States.
Chicago: UnivPrsity oi Chicago Press, 1963.
Machlup, Fritz. "Marginal Analysis and Empitical Research." Am,:rican
Econcmi(' Revie.w 36 (September 1946): 519-554.
."Maignalism, Minimal Wages, and Labor Marl;ets: a &ejoinder
to ;'it Antimarginalist." American Economic Review 37 (March 194?):
148-153.
Marglin, Sto.phan. "What Do Bo3:,es lo?" Peview of Radi.-:.a1 Political
oonomio.s 6 (c17:.1.mer ''orhceming.
Mtse:s, St.xiley H. "rhe E,tect. of :,amily Income on Children's
bducat1on: Findings oa Inequality of Opportunity." Journal
4 Spin; 19(7)1: 158-175.
Miche1s,m, Stopttan. -Income:, of Minorities. Washington, D.C.
1968 (!aimeogr;-Thed).
Mill, J.S. The of 2oliticiil Econmv with of Their
So::: 0 IiiloscnEv. En C,livcted Works
Joirn -_y,1:. Nil]. Yoranto: Univil..rs1ty of Toronto Press, 1965.
or, HermAn P. "Ahnuil .1nd Lifetime Income in Relation to Education."
Amerieln rclnemie. v ow 50 (December 1960) : 962-986.
Mincer, Jacob. "On the Job Tra;nim:: Ces's, Returns, and Some
Implications." journak_oLPolitical yconolor 70 (October 1962):
50-79.
LIZs
411
1,11 [Icor, Jacob. "The D is t_ rihu I or f 1.aitor I ncome!t: A Stirvey." 101naof Econor. ie I. i tera Lure 8 (March 1970) :, 1-26.
Mitchell, Wo:tley C. j'y2esof Jbeorv, Vol. I. NOW York:
Augutttus I. ke 1 Icy, 1967.
Morgan, James N.; David, X.:rtin H.; Cohen, Wilbur J.; and Brazt?x,Harvey E. Incone and Welfare t1,0 ted States.McGraw-ill II , 1962.
New York:
and Si rageld in, 1s;::all. "A Note on die Qealjtv N:s ion itt
Educ:.1 Lion," .1.7urna o 1.itic.jl cons.rav 76 (Sep 1 r tiOcts-ber 1(368)1069-1077.
Mortenien, Dale 1. "A Theory of Wane and Employment :)vnamics." In____Mic roeronon: Foul,t1,.!;_ ion c.f.' ;.:ttolo,,m-ct. and I'nfl a t. toe Theory,
pp. L67-211. Ed I 'oy Lsit.oint_i S. Vto1 ps . Mew York: ';or t on, 1970.
Nichols, Don.ild A. "Nat:1,ot Clear ing for IF:terogeneous Cnp4tal Goods."Mi oe.2enorti ,74 and in flat ion '11;e0u,
pp. 394-410. Ed icet: by Ed:1:1a:: irnelps. Nt_w Yer!:: ;41:.- ton, 1970.
W., Walter. "Labor as a Quasi-Fised Factor."Econo:71cs 70 (Decembr 1962): 538-553.
lournql Political
Quick, Paddy. "Women' =3 Work." R.'.' of Rac.:i cal Pol EconoNics 4(July L972) : 2-19.
Rees, Albert . "Union '.:age Ga , and Eatcrprise non000l y." 1'..sflays onml P t ions Research: Ann Arbor: : iln iversity of Mich igan,_ .
institute of i,,00r and Induri,-.1 Relations, 1961.
and Shull_ 4, George B. Wor1:e rs ond Wnn..--; in !i:: ;:rb.1,1 t al,or.-_, _""o. 1'a u: in hr-,r "it v ol Chicapo ress 1')70
Rensb,t,', "Est i Ian r L rIt i?t '.tt int s to Ecitica Ii cit." Rev i oricc 4 icon 1960) : 31.8-324.
Reynoldi, :Axyd.
Row, 19f.i.ae St rtu.t..1,-e of I..--hor ::ew York: Harper ird
-------
:lad -fa ia H. T I vel tit I, nt of Waoe rut; Lure . New
Haven; `.! ale 1.1: Lye rs t kr Press, 1956.
Rib I a, no,. as I . 1:0ticat_ton and Pover cv. Washi nnton, D.C. : Brook ingsIn =t. tu I ion ,
Po') i nun , A .G. Ind V. , E. , ed s. The Eco ro-,Nm ics at Ed tic a I. ion.1 cittl.;n: St. , 196o.
412
Rose, Gerald. "Differential. Returns to investments in Human Capital
in the Academic Labor Narket." Ph.D. dissertation, University ofCalifornia, ilerk2ley, 1969.
Ross, Arrhur M. TradeUuion Berke ty: University of
California, 1948.
and Ooldnar, William. "F6rces \ffecting the interindustry_______Structure." Ouarterly Je_,..irnal. of Economies f)L (may 1950): 254-261,
Roy, A.D. "fhe Distribution or Earnings and Individual Output."Economic Journal 60 (eptember ).950): 489-505.
Schultz, T.W. "Invest:A:Au_ in Man: An Economist's View." Scp:ial Service
RQVieW 33 (June 1939) : 109-1.17.
"(nvestment in Human Cap11.al." American Economic BLview 51
(March 1)61): 1-17.
Cayo. Education a:1_Inco.ro. New York: Viking Pra:,s,
1964.
Siegel, Paul- "On thc Cost of Being a Negro." Socioloqal locuiryWinter 1965.
Sliel:ter, Sumner. "Notes on the Structure of Wages." R,../Jew of
Pconom1c. 11d ctaristirs 32 (ebew,ry 1950): 80-91.
Sobotka, Swp..en. "11.e Influen,e of Unions on is and l7:arningn of
1,abor i the Construction industry." Ph.D. disertation, University
,.f Cuicago, 1952.
:1orokin, ritirim P. ,11,1itv.. New York: liarpor, 1927.
S affard. F.can': F. "CcnvPrtratioti and label- Earning3: A COTM:11.4
lcon:mL! :eview (March F:68): 17A-181.
)ea.e An 'ncrodoc..ion to QuIntitative E,-onomic_ _ _ _ _ . _ . _ . _ . _ . _(.1::co: Rand M,;,..11.y, 1963.
Sz,rv:,!y of_ r:k !tz:(1 i son: S. E. 0 . ilwiso, Dal a
Comptla: ion Cai. 'r Soi,il Science Buil.iing, The University
of `sc..,n.:in.
Thurw, Lestcr C. Poverty Ind Discriminat ton. Washington, D.C.
:ro.Aings Institute, 1969.
Puma!) tal Beimont: Wadsworth Publ fishing
Comr any, 19i O.
4J0
413
U.S. Congress. Joint EcoaoiiLc Committee. JlleAmorican Distrii)ntion
of lecomc: A Structural Problem, by Lester Thurow and Robert I. B.
Lucaq, JoiaL Co.o.nitreo Print. W:Ishiagton, D.C.: Government
Printing Office, 1972.
. Joint Economic Committee. Polo rat Prof for the
Develou'.:ent of Human Resources. "The Occupltional Distribution
and Returns to Education and Experience for Whites and Negroes,"
by LestLr Thurow, Joint Comittee Print. Was}-int., D.C.:
Goverament Printing Office, 1968.
. Joint Fonomic Committee. Postwar Movement of Prices and
in s2l:alufactrit-0,' inc'e,,tries by Harold M. Levinson, Joint:
Cox.,nittee Pr ott, SZacty P&per 21. Washington, D.C.: Government
Printing Office, 1960.
C. S. Department of Commerce. Brreau of the Census. United States
Census of Population: 1960. "Occupational Characteristics."
Seri PC(2) 7A, 1966.
. BuroNu af the Census. United St.-,tes Census of ropolati.-n:
1360. "industry Chatactetistics. Series PC(2) 7, 1966.
.Current. Popul ttton Renorts--Consumer Income Series P-60,
No. 69, Apr i.1 6, 1970.
.Currcnt Ponn1-(tion Reports: Consumer incom. "Characteristics
of the Low-Income Population 1970." Series P-50, No. 81, 1911.
De.)artmest Labor. ASup-plent. othe Dictionoflocerar.,eanl TitiQS. Washintor., D.C.: U.S. Government Printing
Off,c, 1966.
. Mon:ower Administration. "rho influence of gDTA on Earningu,"
.ower EA,aluaion P.cport No. 8, December 1968.
. Yanp,wer Adis:Istration. "Ir:provements eel. '3 in Cont-ca.:ted
or on-tna-jo:, um:.er ne Manpower Development and
Training Aot." 1968.
. of :rmntion. "Thr,-,e Stat-u:atds of Living for an
!Thal Vaiiv of Four." '.;pring 1967.
' 't.or rii.ace of Pam( I y Heads, by Povert v Status of
mr. t y , 195r;." M-lo2LoweLIZe.pott.. oLt.11e Pres ident. Wash lug ton:
U.S. Government Pri L Ot II(' r, 1970.
Wachtel, Dawn. The 1.:oikin;.: Poor. Mimeographed. Ann Arbor:
in,tituti ot Labor and Indu,trial Relations, University of
Michigan-Wayne State Cniversity, 1967.
431
414
Wachtel., toward and Betsey, Charles. "E.mployment at Low Wages."
lhe Roview of Economics and Statistics 54 (May 1972): 121-129.
Waldman, Elizabvth. "Educationai Attainment of Workers." Monthli
labor Review 9.). (February 1969).
Weber, Max. "Class, Status, and Party." In :lox:Weber. Ess.,:xs in
Sociolop.y. %ew York: Oxfotd University Press, 1946.
Weiss, Leonard W. 'Concentration and Labor Earnings." American
Economic itc.view 56 (:larch 1966): 96-117.
Welch, Finis. "X:,asurt!ment of the Quality of Schooling." American
Ecoomic Rcvi.2w 56 ;May 1966) : 379-392.
Wootton, Bachara. The Sociol Fouudations of Wal:s Pol icy. London:
Litwin Univcrity Books, 1955.
432